Implement operational metrics

[Pokom]

Unlike our cost metrics, operational metrics should be uniform across all providers and collectors. This makes dashboards and alerts a lot easier to setup, e.g. you only have one set of alerts instead of an N*M alerts(where N is providers, and M is collectors). This work has already begun with #104 which can be refactored to belong to the providers package and used by both AWS and GCP. While the current set is a nice foundation, we can extend this even further to:

• cloudcost_exporter_collector_api_requests_total
• cloudcost_exporter_collector_api_requests_errors_total
• cloudcost_exporter_collector_api_requests_duration_seconds
• cloudcost_exporter_collector_last_scrape_time

We'd likely need the following labels:

• provider => CSP
• collector => Module that's making the request
• service => The backend system being called(compute, storage, billing, costexplorer, etc)
• method => The method(ListInstancesInZone, GetServiceName, GetCostUsage)

Once we do this, we can update our existing operational dashboard(https://admin-ops-us-east-0.grafana-ops.net/grafana/d/1a9c0de366458599246184cf0ae8b468/cloudcost-exporter-overview?orgId=1) to use the generic metrics instead of the provider specific ones.

[Pokom]

Just to say, naming is hard and I'm not married to the idea of it being ...collector_api_requests..., I just don't have a better way of communicating external requests.

[Pokom]

This is likely related work to #222 and could follow similar implementations.