fix(cli): emit failure JSON when public DataMap store fails

[jacderida]

Problem

ant --json file upload --public prints no JSON at all when the file body uploads successfully but the trailing public DataMap chunk fails to store. The UploadFailureJson emission only covers the DataError::PartialUpload arm of the body upload; the DataMap path bails straight out via ? (handle_file_upload, file.rs).

Harnesses that parse stdout (e.g. the prod uploader service) see exit≠0 with empty stdout and record 0/0 chunks. Observed on PROD-UL-01 (2026-06-04): 500/500 body chunks stored and paid for, DataMap PUT rejected by 4/7 close-group nodes (see WithAutonomi/ant-node#127), reported as:

│ 1 │ 2015...mp4 │ 1.94 GB │ partial │ 0/0 │ 0 │ 44m 51.6s │

That hides that the body chunks are on the network and paid for — the upload is one 324-byte chunk away from being retrievable, and a DataMap-only retry would be far cheaper than a full re-upload.

Fix

Emit an UploadFailureJson before bailing, matching the PartialUpload arm:

{"error":"datamap_store_failed","total_chunks":501,"chunks_stored":500,"chunks_failed":1,"reason":"failed to store public DataMap: ..."}

Existing consumers need no changes — e.g. the prod uploader's classifier already keys off the error field and reads chunks_stored/reason from the same object.

cargo check -p ant-cli and clippy clean.

Companion PR (root cause of the DataMap rejection itself): WithAutonomi/ant-node#127.

🤖 Generated with Claude Code

[dirvine]

Reviewed the diff locally. This is a small, targeted fix and the JSON failure path now mirrors the existing partial-upload reporting contract for the DataMap-only failure case. Counts look correct: body chunks remain counted as stored and the public DataMap is reported as the single failed chunk.\n\nLocal verification passed:\n- cargo check -p ant-cli\n- cargo clippy -p ant-cli --all-targets -- -D warnings\n- cargo test -p ant-cli\n\nCI is partly still pending, so wait for the remaining build/E2E/Merkle jobs before merging.