fix: Address potential race condition in FeatureStore update_availability

[jsonbailey]

Note

Strengthens availability monitoring and reduces race conditions.

• Refactors FeatureStoreClientWrapper.__update_availability to minimize time under lock, update state atomically, and start/stop RepeatingTask outside the lock; prevents duplicate pollers and adds clearer availability/unavailability warnings
• Tweaks error log in the synchronizer loop to "Failed to build synchronizer" for clearer context

^{Written by Cursor Bugbot for commit 5c5dcd5. This will update automatically on new commits. Configure here.}

[cursor]

Missing try/finally causes deadlock if exception occurs

The refactored __update_availability method removes the try/finally pattern that previously ensured the lock was always released. If any exception occurs between self.__lock.lock() and self.__lock.unlock() (for example, during RepeatingTask construction on line 189), the lock will remain held indefinitely, causing a deadlock for all subsequent operations trying to acquire the same lock. The original code and the equivalent method in ldclient/client.py properly use try/finally to guarantee lock release.