Add `burn archive vacuum` for SQLite maintenance

[willwashburn]

Context

PR #78 shipped burn archive build | rebuild | status but deliberately left vacuum off for the foundation PR:

burn archive vacuum (one-liner; trivial to add when needed).

Issue #40 lists vacuum in the proposed command set:

burn archive vacuum — runs SQLite maintenance if needed.

Over time, archive.sqlite accumulates free pages from INSERT OR REPLACE churn (every stamp re-fold rewrites turn rows; rebuild drops + recreates rows). Without periodic VACUUM, the on-disk file grows beyond what the row count implies. On a developer laptop this is fine for months; on a long-running CI/orchestration host it eventually becomes worth the one-shot reclaim.

Proposal

1. Add a vacuumArchive() function to packages/ledger/src/archive.ts that opens the archive, runs VACUUM, and reports before/after file size.
2. Add a burn archive vacuum subcommand in packages/cli/src/commands/archive.ts that calls it and prints a one-line summary (or JSON via --json):

   archive: vacuumed 12.3 MB -> 4.1 MB (reclaimed 8.2 MB)
   

3. Acquire the same lock the build path uses so we don't VACUUM mid-build.
4. No-op gracefully if the archive doesn't exist (print a hint to run burn archive build first).

Tests:

• Vacuum reduces file size after a build/rebuild churn.
• Concurrent vacuum + build serialize via the lock without corruption.
• Vacuum on a missing archive exits cleanly with a hint.
• --json shape: { before_bytes, after_bytes, reclaimed_bytes }.

Acceptance criteria

• burn archive vacuum exists, documented in command help, and has unit + CLI tests.
• Output (text and --json) includes before/after sizes.
• Concurrent build is safely serialized.
• No-op-with-hint path when archive is missing.

Out of scope

• Auto-vacuum heuristics (e.g. vacuum after N rebuilds). Manual command first; automation later if measurements justify it.
• Schema changes.

Refs

• Spun out of Derived analytics archive foundation: archive.sqlite + burn archive (#40) #78 (Derived analytics archive: materialize the ledger into a local queryable store #40 follow-up)