fix(worker): prevent scheduled workspace purge from failing on overlapping runs

[viethung0o0]

Summary

The hourly workspace-purge cron was surfacing repeated errors in production when it overlapped with another run (multiple worker instances or a run still in progress). The losing run retried for the full lock window and then threw, turning normal contention into a hard failure.

This change makes the cron treat "another run already holds the lock" as a benign skip instead of an error, and speeds up the per-workspace teardown so a single run finishes well within its window.

Changes

• Skip quietly when another purge already holds the distributed lock, instead of throwing.
• Add a short lock-acquire retry window so a contended run gives up quickly rather than blocking.
• Guard against a second local run starting while one is still in progress.
• Tear down due workspaces with bounded concurrency so a large batch completes faster.

Test plan

• redis unit tests (distributed-store.exists)
• business workspace service tests
• worker purge-workspaces handler tests
• pnpm lint

Note: the repo-wide type-check hook currently fails on unrelated, pre-existing errors in ui / database / unbuilt imports package; the packages changed here type-check clean.