fix: coalesce shell progress updates

[kvyb]

Summary

• keep draining and truncating every shell output chunk immediately
• publish changed shell metadata previews at most once per 100 ms
• force the latest changed preview after process settlement

Why

Each metadata callback persists and broadcasts another complete message.part.updated.1 snapshot. Fast or chatty commands can therefore generate many full-preview updates without improving the UI at chunk-level cadence.

A synthetic burst wrote 30 numbered chunks 10 ms apart, then kept the process alive until metadata exposed the final line. Across 10 fresh test processes per implementation:

Implementation	Median metadata updates	Range
Exact baseline	30	30-30
This change	4	4-5

That is an 86.7% median callback-count reduction in this workload. It does not claim an RSS reduction or equivalent production database-byte savings.

The implementation does not group the output stream: pipe draining, output capture, truncation, abort, and timeout behavior still process every chunk. The periodic publisher is scope-bound, deduplicates unchanged previews, and the final forced update prevents a burst ending between intervals from becoming stale.

Related: #33356

Validation

• bun test ./test/tool/shell.test.ts: 25 pass, 0 fail on clean rerun
• bun test ./test/session/prompt.test.ts -t 'cancel finalizes interrupted bash tool output through normal truncation': 1 pass, 0 fail
• bun typecheck: pass
• targeted oxlint on both changed files: 0 errors
• independent defect-first review: no implementation blocker

The package-wide run reached 3,394 pass, 22 skip, 1 todo, with 2 unrelated PTY/session-route failures; both passed immediate targeted reruns.

[github-actions]

This PR doesn't fully meet our contributing guidelines and PR template.

What needs to be fixed:

• PR description is missing required template sections. Please use the PR template.

Please edit this PR description to address the above within 2 hours, or it will be automatically closed.

If you believe this was flagged incorrectly, please let a maintainer know.

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

This pull request has been automatically closed because it was not updated to meet our contributing guidelines within the 2-hour window.

Feel free to open a new pull request that follows our guidelines.

[kvyb]

The closure was correct: I missed the required PR template and closing issue link. I also later found #37653, which already implements the same targeted shell batching fix and was closed only by automated age/reaction cleanup. I will not resubmit this duplicate; I added the new measurements and a reopen request to #37653.