Fix Antares append and unmount FUSE races

[Ivanbeethoven]

Summary

• disable Antares writeback cache to avoid EBADF when appending to existing files
• add a FUSE regression test covering reopen-in-append mode on an existing file
• add a configurable unmount grace period before Antares tears down active mounts

Verification

• cargo test --test antares_test test_fuse_append_existing_file -- --exact --ignored --nocapture --test-threads=1
• cargo test unmount_grace_duration
• manual Antares mount regression for append and unmount-drain behavior

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 900309f0ab

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Reserve the job id before quiescing unmounts

When an unmount is sleeping here, the instances lock has already been dropped, so a concurrent mount_job_at for the same job_id can finish mounting a different custom mountpoint and overwrite both maps before this method reaches fuse_handles.remove(job_id). In that race this unmount removes and tears down the newly inserted AntaresFuse while the original FUSE session can be left orphaned and untracked; the old implementation held the state lock across the unmount and prevented that interleaving. Please reserve/remove or mark the job as unmounting before awaiting the grace period.

Useful? React with 👍 / 👎.