ci: add Claude PR reviewer

[moose-lab]

What

Adds `anthropics/claude-code-action@v1` as an informational PR check.

• Runs on `pull_request` (opened/synchronize/reopened) and on `issue_comment` containing `@claude` (for manual re-trigger).
• Concurrency group per PR, cancels superseded runs.
• Permissions: `contents: read`, `pull-requests: write`, `issues: write` — comment only, no code-write.
• Prompt tuned for this repo: prioritizes correctness, cross-platform safety (incl. TZ — we already have a known coupling in `cost-tracker`), security, TS safety, test coverage. Outputs tagged bullets and a verdict.
• Not added to required status checks on `main` — informational only. A failed Claude run will not block merges.

Pre-merge prerequisites (manual, repo admin only)

This PR is open as a draft until both are done:

1. Install the Claude Code GitHub App on `moose-lab/DevLog`:
   https://github.com/apps/claude
2. Add the `ANTHROPIC_API_KEY` repo secret from your own terminal:
   ```
   gh secret set ANTHROPIC_API_KEY --repo moose-lab/DevLog
   ```
   Get the key from console.anthropic.com → API keys.

Until both are in place, the `Claude Review` check on this very PR will fail
with an auth error — that's expected and is how we'll verify the setup once
the prereqs land. Push an empty commit (or close/reopen) to retrigger.

Why direct Anthropic instead of TokenBoss

TokenBoss's published skill.md explicitly states it only exposes an
OpenAI-compatible Chat Completions endpoint and does not support
Anthropic-native `/v1/messages`. `claude-code-action` requires
`/v1/messages`, so TokenBoss cannot back this action.