Iceberg: clarify trimmed-coverage comments per review

Address Gera's review nits on #14937:
- __init__.py: the distributed transform lists are bound to test params by
  name, so the tuple order must match the parametrize arg-string order, not the
  function-signature order; reword the comment accordingly.
- delete/merge/update _full_coverage tests: refresh the now-stale "full
  partition coverage" docstrings to the sanity-check wording used by the
  CTAS/RTAS/overwrite variants, pointing at the append anchor
  (test_insert_into_partitioned_table_full_coverage).

Signed-off-by: Chong Gao <res_life@163.com>

Author: Chong Gao

integration_tests/src/main/python/iceberg/__init__.py

@@ -119,9 +119,10 @@
 ]
 
 # Each list pairs (partition_col_sql, <mode>). Always used together as
-# `@pytest.mark.parametrize("partition_col_sql,<mode_arg>", <list>)` so the
-# parametrize-string order matches the tuple order and the function-signature
-# order (`partition_col_sql` first, then the mode arg). Keep them in lockstep.
+# `@pytest.mark.parametrize("partition_col_sql,<mode_arg>", <list>)`, so the
+# tuple order must match the order of names in the parametrize arg string
+# (`partition_col_sql` first, then the mode arg). Pytest binds parameters to the
+# test function by name, so the function-signature order does not matter.
 #
 # Each list is just two cases (one per mode, exercising one transform family):
 # the 26-transform partition-writer matrix lives in the append anchor

integration_tests/src/main/python/iceberg/iceberg_delete_test.py

@@ -158,7 +158,10 @@ def test_iceberg_delete_partitioned_table(spark_tmp_table_factory, partition_col
 @pytest.mark.parametrize("partition_col_sql,delete_mode", delete_partition_transforms_distributed)
 @allow_non_gpu_conditional(is_spark_400_or_later(), "EmptyRelationExec")
 def test_iceberg_delete_partitioned_table_full_coverage(spark_tmp_table_factory, partition_col_sql, delete_mode):
-    """Full partition coverage test - skipped for remote catalogs."""
+    """Sanity-check DELETE across the two write modes against partition transforms
+    distinct from those picked by other DML ops. The 26-transform partition-writer
+    coverage anchor lives in
+    iceberg_append_test.py::test_insert_into_partitioned_table_full_coverage."""
     _do_test_iceberg_delete_partitioned_table(spark_tmp_table_factory, partition_col_sql, delete_mode)
 
 @iceberg

integration_tests/src/main/python/iceberg/iceberg_merge_test.py

@@ -183,7 +183,10 @@ def test_iceberg_merge(spark_tmp_table_factory, partition_col_sql, merge_mode):
 @pytest.mark.parametrize("partition_col_sql,merge_mode", merge_partition_transforms_distributed)
 @allow_non_gpu_conditional(is_spark_400_or_later(), "EmptyRelationExec")
 def test_iceberg_merge_full_coverage(spark_tmp_table_factory, partition_col_sql, merge_mode):
-    """Full partition coverage test - skipped for remote catalogs."""
+    """Sanity-check MERGE across the two write modes against partition transforms
+    distinct from those picked by other DML ops. The 26-transform partition-writer
+    coverage anchor lives in
+    iceberg_append_test.py::test_insert_into_partitioned_table_full_coverage."""
     _do_test_iceberg_merge(spark_tmp_table_factory, partition_col_sql, merge_mode)
 
 

integration_tests/src/main/python/iceberg/iceberg_update_test.py

@@ -171,7 +171,10 @@ def test_iceberg_update_partitioned_table_single_column(spark_tmp_table_factory,
 @pytest.mark.parametrize("partition_col_sql,update_mode", update_partition_transforms_distributed)
 @allow_non_gpu_conditional(is_spark_400_or_later(), "EmptyRelationExec")
 def test_iceberg_update_partitioned_table_single_column_full_coverage(spark_tmp_table_factory, update_mode, partition_col_sql):
-    """Full partition coverage test - skipped for remote catalogs."""
+    """Sanity-check UPDATE across the two write modes against partition transforms
+    distinct from those picked by other DML ops. The 26-transform partition-writer
+    coverage anchor lives in
+    iceberg_append_test.py::test_insert_into_partitioned_table_full_coverage."""
     _do_test_iceberg_update_partitioned_table_single_column(spark_tmp_table_factory, update_mode, partition_col_sql)
 
 @iceberg