GenerationSynchronizer is cache-dependent, and when the cache was

down, it would noisily crash. As this was often used in middleware, this

could prevent accessing certain URLs, such as the administration UI and

the new HealthCheckView.
We now protect against this by wrapping all cache-related calls in

exception handlers, logging the errors with some useful contextual

information and informing the admins that the cache server may be down.

This is a strict improvement over the seemingly-random crash behavior

that would manifest before.
Exception handlers are mainly placed in the outer functions where we

know what higher-level operation the synchronizer is trying to perform,

in order to help log better errors.

Tested normal operations, and tested with the cache server down and

in true outage mode. Verified I could still access the health check

endpoint and see useful log entries about the Site Configuration,

extensions, and integrations keys failing to synchronize.