pkgs/id: stale runtime .identity.db databases tracked in git (4 + 22 local identity rows); successor .identity.redb not gitignored #372

Problem

Two stale runtime identity databases are tracked in git:

• pkgs/id/.identity.db — SQLite 3 database, identities table with 4 rows
• pkgs/id/e2e/.identity.db — SQLite 3 database, identities table with 22 rows

Evidence (verified on main a00dd292)

• git ls-files -s pkgs/id/.identity.db pkgs/id/e2e/.identity.db — both tracked as regular files.
• Schema is runtime state, not a fixture: identities(client_id TEXT, name TEXT, created_at INTEGER, updated_at INTEGER) — locally registered identities accumulated by running the server.
• No current code reads either file. git grep -F ".identity.db" finds only a design ledger (thoughts/ledgers/CONTINUITY_ses_2b82.md) describing the old libsql design: "Store in encrypted Turso/libsql DB (.identity.db next to .iroh-key)".
• The current code uses a different filename: pkgs/id/src/commands/serve.rs:1098 constructs .identity.redb (Path::new(KEY_FILE).with_file_name(".identity.redb")).
• The e2e suite does not use it as a seed/fixture: no reference in pkgs/id/e2e/ (Playwright config starts ephemeral servers: ../target/debug/id serve --web --ephemeral).
• Nothing ignores the pattern: pkgs/id/.gitignore contains only .iroh* and test.*; git check-ignore pkgs/id/.identity.db returns nothing. The neighbouring runtime key material is ignored via .iroh*, but the identity DB escapes that pattern — and the current .identity.redb runtime file is not ignored either, so it is one accidental git add -A away from the same problem.
• History: the files arrived in bulk April commits (c0a30175 "idk stuff from april").

This is the same class as #367 (.opencode/ runtime state) and #370 (.claude/settings.local.json), but a distinct location those issues do not cover. Single-repo: no .identity.db is tracked in developing-today/opencode, developing-today/openchamber, or dezren39/nix (checked git ls-files in all three).

Why it matters

• Local identity registrations (names/client IDs and timestamps from whoever ran the server) leak into the shared repo and go stale immediately.
• A stale DB checked out fresh can be mistaken for a seed/fixture, and if the old libsql path is ever revived it would silently load April's identities.
• The un-ignored .identity.redb successor is the live version of the same hazard.

Acceptance criteria

• git ls-files pkgs/id/.identity.db pkgs/id/e2e/.identity.db is empty (both untracked).
• git check-ignore matches .identity.db and .identity.redb under pkgs/id/ (including the e2e/ subdirectory) so neither runtime DB can be re-committed.
• No code, test, or e2e configuration references either tracked file (already verified — please re-check before removing).
• Local copies may stay on disk for whoever uses them; this is an untrack + gitignore change only, like chore: .claude/settings.local.json is tracked in git — personal local settings with machine-specific paths and a broad home-read allow #370 / fix(lab-ops-hub): judge NAT64/6to4/mapped crawl literals by embedded IPv4 #473.