Add critical hardware tests for real-world breakage scenarios

Track previously untracked test support files

Expand test suite to ~170 tests across 3 tiers

Fix run_on_robot.py: create remote dir before rsync, allow password auth

Add smoke test to verify full playback pipeline on robot

Add remote hardware testing via SSH and redesign motion accuracy tests

Fix hardware test cascade: isolate audio tests, skip on mic failure

Add deep pipeline integration tests for hardware suite

Fix recording freeze: handle queued mode in stop endpoint

Add audio playback test to hardware suite

Improve testing: verbose output, web coverage panel, hardware tests

Add test documentation, runner script, and coverage matrix

Fix HF test: allow real username when huggingface_hub is installed

Separate local and downloaded datasets: block recording into downloads

Add HF auto-login detection: auto-fill username from CLI session

Add 'starting_up' mode for non-blocking startup animation

Remove denoise feature from UI and backend

Simplify recording form: Name + Duration + Record in one row

Fix P0-1: Replace fragile progress bar timing with server-driven phases

Fix P0-2: Allow non-0.5-second durations in recording form

Add test infrastructure: 50 unit tests + 13 Playwright E2E tests