refactor: simplify PhasedCodecPipeline — remove layout abstraction, use codec chain directly

Remove ~840 lines of ChunkLayout hierarchy (ShardIndex, SimpleChunkLayout,
ShardedChunkLayout, fetch_chunks_sync/async, decode_chunks_from_index,
merge_and_encode_from_index). The pipeline now uses ChunkTransform directly
for sync decode/encode and falls back to the async codec API otherwise.

Also fix ShardingCodec._encode_sync to respect write_empty_chunks config
by skipping inner chunks that are all fill_value.

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

Author: d-v-b

src/zarr/codecs/sharding.py

@@ -536,10 +536,19 @@ def _encode_sync(
             morton_order_iter(chunks_per_shard)
         )
 
+        chunk_spec = self._get_chunk_spec(shard_spec)
+        skip_empty = not shard_spec.config.write_empty_chunks
+        fill_value = shard_spec.fill_value
+        if fill_value is None:
+            fill_value = shard_spec.dtype.default_scalar()
+
         for chunk_coords, _chunk_selection, out_selection, _ in indexer:
             chunk_array = shard_array[out_selection]
-            encoded = inner_transform.encode_chunk(chunk_array)
-            shard_builder[chunk_coords] = encoded
+            if skip_empty and chunk_array.all_equal(fill_value):
+                shard_builder[chunk_coords] = None
+            else:
+                encoded = inner_transform.encode_chunk(chunk_array)
+                shard_builder[chunk_coords] = encoded
 
         return self._encode_shard_dict_sync(
             shard_builder,