Expand description
The same bulk-bytes treatment for a plain Vec<u8> field.
Used by cross-machine payloads that do not need the 128-byte Arrow
alignment (InterDaemonEvent::MemoryPoolWrite::tensor_data), where the
receiver copies the bytes into a mirror segment rather than decoding them
zero-copy. The wire encoding is identical to serde’s default Vec<u8>
(postcard: varint length + raw bytes), so this is a pure speedup with no
version bump — vec_encoding_is_unchanged_from_the_seq_form pins it.