fix: update stale test referencing removed _transform_read method

Test now uses decode_chunks_from_index with ShardIndex.leaf_transform
instead of the deleted pipeline._transform_read.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: d-v-b

tests/test_phased_codec_pipeline.py

@@ -319,8 +319,12 @@ def test_simple_layout_decode_skips_indexer() -> None:
     encoded = pipeline.layout.inner_transform.encode_chunk(data)
     assert encoded is not None
 
-    # Decode via _transform_read — should use fast path
-    result = pipeline._transform_read(encoded, spec)
+    # Decode via decode_chunks_from_index using leaf_transform
+    from zarr.core.codec_pipeline import ShardIndex, decode_chunks_from_index
+
+    index = pipeline.layout.resolve_index(None, "c/0")
+    raw_chunks = {(0,): encoded}
+    result = decode_chunks_from_index(raw_chunks, index, spec)
     assert result is not None
     np.testing.assert_array_equal(result.as_numpy_array(), np.arange(100, dtype="float64"))