refactor(hooks): split config and executor modules

[cyq1017]

Problem

crates/tui/src/hooks.rs mixed hook config definitions with executor runtime behavior in one large module, which made hooks policy changes harder to review.

Change

• Move hook events, conditions, hook definitions, and HooksConfig into hooks/config.rs.
• Move HookExecutor, subprocess execution, timeout handling, stdout parsing, payload helpers, and tests into hooks/executor.rs.
• Keep the existing root crate::hooks::* API available through re-exports.

Closes #4082.

Verification

• cargo test -p codewhale-tui hooks:: --locked -- --nocapture
• RUSTFLAGS='-Dwarnings' cargo test -p codewhale-tui hooks:: --locked -- --nocapture
• cargo fmt --all -- --check
• git diff --check

[Hmbown]

Thanks — this is the split #4082 asks for, done cleanly: hooks/config.rs holds the event/matcher/HooksConfig types with no subprocess dependencies, hooks/executor.rs owns HookExecutor, spawn/wait/timeout, and stdout-decision parsing along with the 56 moved tests, and the nine-line hooks.rs shim keeps crate::hooks::* stable. The head also has a fully green matrix (ubuntu/macos/windows tests, lint, DCO) from July 16 — appreciated.

Where it stands: #4082 is on the v0.9.3 milestone, behind the v0.9.1 freeze and the active v0.9.2 lane, so there is runway — but the branch is stale in the one way that matters for a file move. Main's hooks.rs picked up three fixes on July 17-18 (ee8b6db process-tree containment, b8258fc observer EOF via closed pipe, 525ef8a Windows EOF regression test) and now sits at 3,018 lines versus 2,593 at the cut, so the move will conflict and a naive resolution could drop those fixes.

Path to land:

1. Regenerate the split on current main rather than doing conflict surgery — re-apply the same module boundaries so the new executor fixes end up inside hooks/executor.rs.
2. Re-run cargo fmt and cargo test -p codewhale-tui hooks::, and note in the body that the v0.9.3 refactor(hooks): split config types from HookExecutor #4082 acceptance criteria hold (config module free of subprocess imports; decision-JSON and timeout tests unchanged).
3. Mark the PR ready for review.

If you would rather not redo the move, say so — it is small enough to regenerate maintainer-side and land with your credit (Co-authored-by plus Harvested from PR #4087 by @cyq1017). Closing #4082 stays attached to your work either way.

---

Generated by Claude Code