Refactor the metric names for spec to be subsystem prefixed with 'spec'.
expose metrics for spec loading and failures to parse.
expose metrics for PKI writes to disk, and failure counts.
export metric count for all remote request failures, whether it be CA fetching
or cert signing.
Being we now can force regeneration for hostname/algo/key size
changes, drop the metrics that track this. If we need to track this
level of granualarity, the logs have it- but now that we regenerate
we don't need to surface this to prometheus.
Wipe metrics on spec reload to prevent stale metrics.
Fix updateCAExpiry to set the correct metric label value
export a metric per spec for the 'before' threshold that is internally used.
rough list of changes: