fix: fail closed unresolved watch chats

[steipete]

Summary

• retry watch rows when all-chat polling sees a message before chat_message_join metadata exists
• stop the retry loop from advancing past that unresolved row until it resolves or exhausts the retry budget
• drop persistently unresolved rows fail-closed instead of emitting chat_id=0 DM-shaped payloads

Fixes #118.

Proof

• swift test --filter 'MessageWatcherTests|RPCServerTests'
• make lint (passes; existing warning-only lint debt remains)
• /Users/steipete/Projects/agent-scripts/skills/autoreview/scripts/autoreview --mode local --parallel-tests "swift test --filter 'MessageWatcherTests|RPCServerTests'" clean

[clawsweeper]

ClawSweeper status: review started.

I am starting a fresh review of this pull request: fix: fail closed unresolved watch chats This is item 1/1 in the current shard. Shard 0/1.

This placeholder means the worker is alive and reading the current context. I will edit this same comment with the actual review when the claws are done clicking.

Crustacean status: shell secured, claws on keyboard, evidence pebbles being sorted.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c18707a686

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]

Avoid exhausting retries with zero debounce

When callers configure no debounce, this retry loop becomes effectively immediate: watchDebounceIntervalParam accepts debounce_ms values >= 0, and the CLI parser also accepts --debounce 0ms, so schedulePoll() requeues another poll with no delay until the 20-attempt budget is spent. In that configuration a row whose chat_message_join appears a few milliseconds later is dropped fail-closed before Messages has time to resolve the chat metadata, which reintroduces missed live messages for low-latency watch/RPC users; use a minimum retry delay or wall-clock retry window independent of the event debounce.

Useful? React with 👍 / 👎.