notifications metrics code clean up

The current metrics code could use some cleanup and refactoring in order to make adding new metrics easier and more robust:

• prometheus metrics do not require locking
• unify how we define and test metrics with e.g. storage metrics
• add tests for metrics
• make notifications metrics available through /debug endoint lock-free - there is no need for global lock
• add destingation endpoing labels to pending and status metrics
• add some basic documentation