fix(test): temp dir cleanup + testTimeout for parallel workers

[diazMelgarejo]

While working on the main-merged old MacOS PR and testing it, we encountered these stalling files (changes from the most recent upstream fixes, especially v0.9.9 with Commit 092df06). I decided to separate this from the other commits for MacOS port.

Root cause (two layers):

1. routes-browse.test.js + routes-models.test.js called mkdtempSync per test but never cleaned up. 30+ leaked dirs/run add I/O overhead on macOS /var/folders/ (memory-mapped filesystem).
2. db-layer parallel workers (doctor, watchdog, webhooks) leave DatabaseSync .db-shm pages mmap'd. Under 60+ workers on ARM64, OS memory pressure slows ALL workers — including pure-mock routes tests — past the 5s timeout.

Fix:

• routes-browse.test.js: track created dirs in createdTestRoots[]; afterEach splices and rmSync({recursive,force}) each entry
• routes-models.test.js: same pattern for createApp() temp dirs (createdTempRoots)
• vitest.config.js: testTimeout 5000 → 10000ms; gives pure-mock workers breathing room while db-layer .db-shm pressure fix propagates (doctor/watchdog/webhooks-db still need afterEach closeXyzDb() — tracked as remaining upstream work)

Verified: npm run test:coverage passes 594/594 sequentially.
Target: npm test 594/594 (was 584/594 before this fix).