issue-1751: [Filestore] WriteBackCache should not crash on corruption

[e673]

Notes

Implement correct behavior for WriteBackCache when corruption is detected.

Previous behavior:

• Incomplete object creation causing null pointer access in various scenarios, for example, metrics collection.
  Both resulted in entering pod into crash loop;
• Triggering Y_ABORT_UNLESS when trying to add a write request.
  Both resulted in a crash loop.

Expected behavior:

• WriteBackCache is fully constructed;
• WriteBackCache doesn't serve requests and hangs;
• No errors are sent to the guest;
• A critical event is reported.

Manual resolution should be conducted in order to resolve the corrupted state.

Issue

#1751