Refactor RepairState, metrics and alarms

[masokol]

We should do some refactoring regarding schedules and repairState. Currently metrics and alarms depend on when the repairState changes snapshot, i.e if it changes too fast/too slow ecChronos either ~spams~ alarms, or don't do alarms, metrics might be outdated/not present as stated in #429.

Proposal:

• RepairSnapshotProvider (Caches repairStates/repairStateSnapshots and updates them periodically) - Used by schedules to fetch snapshot. The frequency of repairSnapshot updates should be exposed in ecc.yml