fix(otel_metrics): use opentelemetry:timestamp/0

[SergeTupchiy]

There was inconsistency in timestamps handling: metrics used system_time(nanosecond), while metrics exporter treated these values as opentelemetry:timestamp() (monotonic time in native units).

This PR fixes it by using opentelemetry:timestamp(), most of the diff is caused by renaming *time_unix_nano to time to avoid confusions.

[SergeTupchiy]

@tsloughter , @bryannaegele, @ferd
Could you please take a look at this PR?
Alternative fix would be to avoid converting timestamps in otel_otlp_metrics: https://github.com/open-telemetry/opentelemetry-erlang/blob/main/apps/opentelemetry_experimental/src/otel_otlp_metrics.erl#L89
But I think the proposed fix is better as it follows the same approach used for traces: https://github.com/open-telemetry/opentelemetry-erlang/blob/main/apps/opentelemetry_api/src/opentelemetry.erl#L305

[bryannaegele]

This seems good? I haven't been involved in the metrics but this seems cleaner and more consistent.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

see 55 files with indirect coverage changes

📢 Thoughts on this report? Let us know!.

[tsloughter]

Thanks, yea, I think this looks right.

The only thing I'm debating, which can be a separate issue/pr, is should we really be using monotonic_time instead of just system_time... I don't think we ever use opentelemetry:timestamp anywhere it gets used to compare or find a diff. Maybe safest to allow it to be used for comparisons/diffs.