Resiliency: Infrastructure Backoff & Best-Effort Release

Nobody has claimed this yet.

### Description
Fortify the LockCoordinator with infrastructure backoffs and best-effort cleanup logic. If the config-server returns 5xx infrastructure errors or network timeouts during lock acquisition, the SDK must back off exponentially (e.g., 1s, 2s, 4s, 8s) to protect the backend. Additionally, the final releaseLock execution within the cleanup block must silently swallow network errors, ensuring that transient cleanup failures do not crash the successfully updated Node process.

### Acceptance criteria
- [ ] Simulating a 5xx error or network timeout during lock acquisition triggers exponential backoff (verified via undici MockAgent and vi.useFakeTimers()).

- [ ] Simulating a network error during the releaseLock call successfully catches and swallows the error without throwing an unhandled exception to the main Node process.

### Blocked by
Issue #126

User stories covered: 8, 9

The work centers on LockCoordinator lock acquisition and the cleanup-block releaseLock call; it is blocked by issue #126. Start by locating those entry points, then use undici MockAgent with vi.useFakeTimers() to verify exponential backoff for 5xx and timeout failures, and confirm releaseLock network errors are swallowed without an unhandled exception.

Written by the indexing model from the issue text.