Buffered upload drops a chunk after transport-level PUT failure

Unconfirmed. This is the result of a search with Claude, may be a false positive.

During a buffered upload, a connection reset, timeout, or other transport error during a chunk PUT returns an `Err` from `send().await`:

https://github.com/googleapis/google-cloud-rust/blob/f02b42c2ac5720334cd2b5335824bd4166264527/src/storage/src/storage/perform_upload/buffered.rs#L127

Because the error is propagated immediately using `?`, `progress.handle_error()` is never invoked. On the subsequent retry attempt, `progress.needs_query()` remains `false` because `persisted_size` still matches `offset`:

https://github.com/googleapis/google-cloud-rust/blob/f02b42c2ac5720334cd2b5335824bd4166264527/src/storage/src/storage/perform_upload/buffered/progress.rs#L104-L106

Consequently, the status query is skipped and `next_buffer()` replaces the unacknowledged chunk `k` with chunk `k + 1`, which is then PUT at chunk `k`'s offset. Because checksums are not sent on resumed chunks, the server accepts the corrupted/truncated upload.

Start by reading src/storage/src/storage/perform_upload/buffered.rs around the chunk PUT and src/storage/src/storage/perform_upload/buffered/progress.rs around needs_query(). Confirm the transport-error retry path and add coverage for a failed chunk PUT; done means the unacknowledged chunk is not replaced and the resumed upload remains correct.

Written by the indexing model from the issue text.