The API goldens now include dunder methods defined in TensorFlow

This allows using the goldens to generate more accurate type stubs, where
e.g. `tf.Tensor` supports binary operators and `tf.named_scope` is a context
manager.

I had to also fix two issues uncovered by this change

* `tf.Tensor` dunder methods don't always use `self`, so the name-based
  `staticmethod` heuristic added earlier misfired. I moved
  classmethod/staticmethod detection to the API generator, where we
  have access to the runtime objects.
* pybind11 docstring parsing didn't support *args, which led to
  `__exit__` having the wrong signature for some types.
* `__round__` is monkey-patched into `tf.Tensor` at import time by NumPy
  "behavior" internals, so we exclude it from the API to keep the API
  generation independent from import side-effects.

PiperOrigin-RevId: 948201041

Author: superbobry

tf_keras/api/golden/v1/tensorflow.keras.-model.pbtxt

@@ -164,6 +164,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\'], varargs=args, keywords=kwargs, defaults=None"
@@ -259,6 +263,7 @@ tf_class {
   member_method {
     name: "from_config"
     argspec: "args=[\'cls\', \'config\', \'custom_objects\'], varargs=None, keywords=None, defaults=[\'None\'], "
+    method_kind: CLASS
   }
   member_method {
     name: "get_build_config"

tf_keras/api/golden/v1/tensorflow.keras.-sequential.pbtxt

@@ -166,6 +166,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\', \'layers\', \'name\'], varargs=None, keywords=None, defaults=[\'None\', \'None\'], "
@@ -265,6 +269,7 @@ tf_class {
   member_method {
     name: "from_config"
     argspec: "args=[\'cls\', \'config\', \'custom_objects\'], varargs=None, keywords=None, defaults=[\'None\'], "
+    method_kind: CLASS
   }
   member_method {
     name: "get_build_config"

tf_keras/api/golden/v1/tensorflow.keras.__internal__.layers.-base-random-layer.pbtxt

@@ -127,6 +127,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\', \'seed\', \'force_generator\', \'rng_type\'], varargs=None, keywords=kwargs, defaults=[\'None\', \'False\', \'None\'], "

tf_keras/api/golden/v1/tensorflow.keras.__internal__.legacy.layers.-average-pooling1-d.pbtxt

@@ -139,6 +139,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\', \'inputs\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\', \'pool_size\', \'strides\', \'padding\', \'data_format\', \'name\'], varargs=None, keywords=kwargs, defaults=[\'valid\', \'channels_last\', \'None\'], "

tf_keras/api/golden/v1/tensorflow.keras.__internal__.legacy.layers.-average-pooling2-d.pbtxt

@@ -139,6 +139,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\', \'inputs\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\', \'pool_size\', \'strides\', \'padding\', \'data_format\', \'name\'], varargs=None, keywords=kwargs, defaults=[\'valid\', \'channels_last\', \'None\'], "

tf_keras/api/golden/v1/tensorflow.keras.__internal__.legacy.layers.-average-pooling3-d.pbtxt

@@ -139,6 +139,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\', \'inputs\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\', \'pool_size\', \'strides\', \'padding\', \'data_format\', \'name\'], varargs=None, keywords=kwargs, defaults=[\'valid\', \'channels_last\', \'None\'], "

tf_keras/api/golden/v1/tensorflow.keras.__internal__.legacy.layers.-batch-normalization.pbtxt

@@ -139,6 +139,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\', \'inputs\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\', \'axis\', \'momentum\', \'epsilon\', \'center\', \'scale\', \'beta_initializer\', \'gamma_initializer\', \'moving_mean_initializer\', \'moving_variance_initializer\', \'beta_regularizer\', \'gamma_regularizer\', \'beta_constraint\', \'gamma_constraint\', \'renorm\', \'renorm_clipping\', \'renorm_momentum\', \'fused\', \'trainable\', \'virtual_batch_size\', \'adjustment\', \'name\'], varargs=None, keywords=kwargs, defaults=[\'-1\', \'0.99\', \'0.001\', \'True\', \'True\', \'<tensorflow.python.ops.init_ops.Zeros object instance>\', \'<tensorflow.python.ops.init_ops.Ones object instance>\', \'<tensorflow.python.ops.init_ops.Zeros object instance>\', \'<tensorflow.python.ops.init_ops.Ones object instance>\', \'None\', \'None\', \'None\', \'None\', \'False\', \'None\', \'0.99\', \'None\', \'True\', \'None\', \'None\', \'None\'], "

tf_keras/api/golden/v1/tensorflow.keras.__internal__.legacy.layers.-conv1-d.pbtxt

@@ -139,6 +139,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\', \'inputs\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\', \'filters\', \'kernel_size\', \'strides\', \'padding\', \'data_format\', \'dilation_rate\', \'activation\', \'use_bias\', \'kernel_initializer\', \'bias_initializer\', \'kernel_regularizer\', \'bias_regularizer\', \'activity_regularizer\', \'kernel_constraint\', \'bias_constraint\', \'trainable\', \'name\'], varargs=None, keywords=kwargs, defaults=[\'1\', \'valid\', \'channels_last\', \'1\', \'None\', \'True\', \'None\', \'<tensorflow.python.ops.init_ops.Zeros object instance>\', \'None\', \'None\', \'None\', \'None\', \'None\', \'True\', \'None\'], "

tf_keras/api/golden/v1/tensorflow.keras.__internal__.legacy.layers.-conv2-d-transpose.pbtxt

@@ -140,6 +140,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\', \'inputs\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\', \'filters\', \'kernel_size\', \'strides\', \'padding\', \'data_format\', \'activation\', \'use_bias\', \'kernel_initializer\', \'bias_initializer\', \'kernel_regularizer\', \'bias_regularizer\', \'activity_regularizer\', \'kernel_constraint\', \'bias_constraint\', \'trainable\', \'name\'], varargs=None, keywords=kwargs, defaults=[\'(1, 1)\', \'valid\', \'channels_last\', \'None\', \'True\', \'None\', \'<tensorflow.python.ops.init_ops.Zeros object instance>\', \'None\', \'None\', \'None\', \'None\', \'None\', \'True\', \'None\'], "

tf_keras/api/golden/v1/tensorflow.keras.__internal__.legacy.layers.-conv2-d.pbtxt

@@ -139,6 +139,10 @@ tf_class {
     name: "weights"
     mtype: "<class \'property\'>"
   }
+  member_method {
+    name: "__call__"
+    argspec: "args=[\'self\', \'inputs\'], varargs=args, keywords=kwargs, defaults=None"
+  }
   member_method {
     name: "__init__"
     argspec: "args=[\'self\', \'filters\', \'kernel_size\', \'strides\', \'padding\', \'data_format\', \'dilation_rate\', \'activation\', \'use_bias\', \'kernel_initializer\', \'bias_initializer\', \'kernel_regularizer\', \'bias_regularizer\', \'activity_regularizer\', \'kernel_constraint\', \'bias_constraint\', \'trainable\', \'name\'], varargs=None, keywords=kwargs, defaults=[\'(1, 1)\', \'valid\', \'channels_last\', \'(1, 1)\', \'None\', \'True\', \'None\', \'<tensorflow.python.ops.init_ops.Zeros object instance>\', \'None\', \'None\', \'None\', \'None\', \'None\', \'True\', \'None\'], "