[Optimize, NPU] Remove tl.where from _rms_norm_forward/backward_kernel_tiled() (#1153)

## Summary

When the mask has a large shape, `tl.where` is not NPU-friendly in
`triton-ascend`, leading to low kernel performance. When writing
kernels, it's best to use alternative logic instead. This can result in
significant performance improvements.

**Will these changes affect accuracy?** Since the masking operation has
already been applied when loading X_block, it will not affect the
calculation result during tl.sum.

## Testing Done

### Accuracy first

The shapes in `test_rms_norm.py` are too small to trigger the
`_rms_norm_forward_kernel_tiled` kernel, so we need a new configuration.

```python
@pytest.mark.parametrize(
    "bs, sl, hd",
    [
        (2, 2048, 4096),
        (2, 2048, 8192),
        (2, 2048, 16384),
        (2, 2048, 32768),
    ],
)
@pytest.mark.parametrize(
    "dtype, atol, rtol",
    [
        (torch.float32, 1e-4, 1e-6),
    ],
)
@pytest.mark.parametrize(
    "reference, offset, casting_mode",
    [
        (LlamaRMSNorm, 0.0, "llama"),
        (GemmaRMSNorm, 1.0, "gemma"),
        pytest.param(
            BaseRMSNorm,
            0.0,
            "none",
            marks=pytest.mark.skipif(device == "npu", reason="Ascend NPU does not support this test"),
        ),
    ],
)
@pytest.mark.parametrize(
    "in_place",
    [
        True,
        False,
    ],
)
@pytest.mark.parametrize(
    "elementwise_affine",
    [
        True,
        False,
    ],
)
```

#### Env
<img width="696" height="251" alt="image"
src="https://github.com/user-attachments/assets/c0cd523f-8a78-4205-8f18-6854477b9d0a"
/>


#### Results after code modification

<img width="1178" height="836" alt="image"
src="https://github.com/user-attachments/assets/b520faa0-f209-4ce0-9574-6000bb612f91"
/>


### Benchmark test

The test cases in `benchmark_rms_norm.py` should keep the same shapes as
those in `test_rms_norm.py`.

```python
common_configs = {
    ...
    "x_values": [2**i for i in range(12, 16)],
    ...
}
```

#### Before Optimization

forward
<img width="1000" height="600" alt="rms_norm_speed_forward"
src="https://github.com/user-attachments/assets/92a701fc-37a9-4298-a285-95174f97ef98"
/>

backward
<img width="1000" height="600" alt="rms_norm_speed_backward"
src="https://github.com/user-attachments/assets/8cfa175d-d6d3-4472-9b9d-7162e6a9df02"
/>

full
<img width="1000" height="600" alt="rms_norm_speed_full"
src="https://github.com/user-attachments/assets/f8248fc6-52a5-4963-a74e-54f7d85173cd"
/>

memory
<img width="1000" height="600" alt="rms_norm_memory_full"
src="https://github.com/user-attachments/assets/7d4b01bb-d425-4e2b-87ef-3d3b1638324b"
/>


[all_benchmark_data_raw.csv](https://github.com/user-attachments/files/26131388/all_benchmark_data_raw.csv)

#### After Optimization

forward
<img width="1000" height="600" alt="rms_norm_speed_forward"
src="https://github.com/user-attachments/assets/89b1a07f-273a-4d07-a6c2-5eb20c7abeb1"
/>

backward
<img width="1000" height="600" alt="rms_norm_speed_backward"
src="https://github.com/user-attachments/assets/4bd741c9-d401-4ade-ac7c-e3b8a241bc32"
/>

full
<img width="1000" height="600" alt="rms_norm_speed_full"
src="https://github.com/user-attachments/assets/c11f7b8c-a3d9-433e-a24e-ebdae343e2e9"
/>

memory
<img width="1000" height="600" alt="rms_norm_memory_full"
src="https://github.com/user-attachments/assets/e2bfedf2-65b2-440d-96fc-4f389cd5ddda"
/>


[all_benchmark_data_optimized.csv](https://github.com/user-attachments/files/26131403/all_benchmark_data_optimized.csv)


- Hardware Type: Atlas 900 A2 PoD
- [x] run `make test` to ensure correctness
- [x] run `make checkstyle` to ensure code style
- [x] run `make test-convergence` to ensure convergence

Author: zheliuyu

src/liger_kernel/ops/backends/_ascend/ops/rms_norm.py

@@ -199,7 +199,7 @@ def _rms_norm_forward_kernel_tiled(
                 X_block = X_block.to(tl.float32)
 
             # Accumulate sum of squares (only for valid elements)
-            sum_square += tl.sum(tl.where(mask, X_block * X_block, 0.0))
+            sum_square += tl.sum(X_block * X_block)
 
         # Compute rstd for this row
         mean_square = sum_square / n_cols
@@ -456,7 +456,7 @@ def _rms_norm_backward_kernel_tiled(
                     m = dY_block
 
             # Accumulate sum(m * X)
-            sum_m_X += tl.sum(tl.where(mask, m * X_block, 0.0))
+            sum_m_X += tl.sum(m * X_block)
 
         # Compute the correction factor
         correction_factor = -(1.0 / n_cols) * rstd * rstd * sum_m_X