Add health metrics for SDK metrics part

[JonasKunz]

Adds health metrics for the SDK metric parts. Based on the span health metrics from #1631 and log health metrics from #1921.

Changes

• Adds a new histogram metric for the number, duration and failures of metric collections per metric-reader
• Adds exporter metrics for metric-exporters: these are pretty much a copy-paste, which is in accordance with Reality-check if exporter health metrics should be merged across signals #2002.

In contrast to for span and log processors, this PR does not include a metric counting the number of metric data points seen by a metric reader. The reason for this is that in the case a Metric Producer fails, there is no way to know how many data points it would have produced.

The metric reader collection duration metric will still show problematic metric producers:

• If a metric producer takes too long, the metric reader collection will also take too long, which will be visible in the metric
• If any metric producer fails, the collection will be marked as failure in the metric

So the metric will be good enough to detect problems, but to investigate the root cause (e.g. finding the offeninding metric producer) one will need to dive into logs.

Merge requirement checklist

• CONTRIBUTING.md guidelines followed.
• Change log entry added, according to the guidelines in When to add a changelog entry.
  • If your PR does not need a change log, start the PR title with [chore]
• schema-next.yaml updated with changes to existing conventions.