Connect bot-sdk's Metrics engine up to Prometheus for finer-grained metrics

It won't be perfect, but it'd allow us to measure internal application health like redaction failures and such. Currently we appear to only expose default NodeJS metrics, which aren't detailed enough to make automated decisions on.

In future it'd be nice to also have metrics for actual internal health exclusive of the bot-sdk operations.

Locate the bot-sdk Metrics engine and the existing default NodeJS metrics entry point, then review how Prometheus metrics are currently exposed. Define which internal health signals, including redaction failures, should be exported and verify that the resulting metrics support finer-grained application-health decisions without removing the existing NodeJS metrics.

Written by the indexing model from the issue text.