Whenever metrics collection fails in the Prometheus agent, two
unexpected behaviors are present:
  - The server does not respond with a 500 error
  - The offending exception is unhandled

Adjust the collection code so that prom-client's exception are correctly
handled. Additionally, make sure to only write response headers _after_
the collection was fully executed; if it fails before then, respond with
a 500 error.

The setCollectorWithGenerator function in the Prometheus agent wrapper
does not handle promisified generator functions correctly. It tries
settings the Promise value returned as the metric value instead of
waiting for resolution.

Wait for generators to resolve before setting the metrics value.

We haven't got a way to monitor Redis HASH/SET leaks because we're not
monitoring them.

Add two new gauges for user and meeting ID mappings stored in Redis:
  - bbb_webhooks_meeting_mappings
  - bbb_webhooks_user_mappings