chore: add a .gitignore that ignores worktree directories

[jaryl]

This repo had no .gitignore at all, so any worktree created here reports the tree dirty — which is how it surfaced during a workspace cleanup.

Two conventions are in use across the estate — .worktrees/ for developer worktrees and .claude/worktrees/ for agent ones — and the repos that were already quiet ignore both. .DS_Store rides along since macOS is the only development platform here.

Matches the same fix just landed in rarebit-one/jumpdrive-runner#64, rarebit-one/jumpdrive-static#66, rarebit-one/rarebit-static-v3#323 and rarebit-one/nutripod-web#1125.

Cosmetic in isolation, but a permanently-dirty tree is corrosive out of proportion to its size: it trains you to stop reading git status, and it makes /cleanup's pristine pass flag this repo every single run — so the one time it flags something real, it reads as the usual noise.

[claude]

Claude finished @jaryl's task in 46s —— View job

---

Code Review

• Gather context (PR diff, CLAUDE.md conventions)
• Review the .gitignore contents
• Post review feedback

Summary: Small, well-scoped .gitignore addition — no issues found.

What I checked:

• Correctness: .worktrees/ matches the convention documented in this repo's CLAUDE.md (git worktree add .worktrees/<name> ...), and .claude/worktrees/ matches the agent-worktree convention referenced in the PR description. Confirmed via git ls-files that no currently-tracked files fall under either path or match .DS_Store — so this won't silently hide anything already in the tree.
• Gitignore semantics: .DS_Store (no leading/trailing slash) correctly matches at any directory depth, so a separate **/.DS_Store entry isn't needed.
• File hygiene: Ends with a trailing newline, no trailing whitespace.
• Security/performance/tests: Not applicable — this is a non-executable ignore-list addition with zero runtime surface.

Nit (optional, not blocking): The comment block explaining the dual .worktrees/ / .claude/worktrees/ convention is genuinely useful here since it's non-obvious from the file alone — good call keeping it, since a bare .gitignore with no rationale would likely just get "cleaned up" by someone later without contacting the estate-wide convention it's protecting.

Matches the stated goal of eliminating the always-dirty tree noise this repo was producing. LGTM.