Remove timestamp read per metric update

[cijothomas]

Changes

Removes the usage of DateTimeOffset in every measurement record event. There seems to be no need of paying cost for reading/passingaround/storing the time with every Update call.

The StartTime is read when the Aggregators start up. (same as before)
The EndTime is ready when "collect" occurs. (same as before)

There can be some more changes done the same area (in future PRs), where we read system time just once at startup, and then rely purely on tick counts, to calculate endtimes for each Collect/Export cycle.

For significant contributions please make sure you have completed the following items:

• CHANGELOG.md updated for non-trivial changes
• Design discussion issue #
• Changes in public API reviewed

[victlu]

It feels funny to not have DT per measurement, but technically, we can leave this to Aggregators that actually need it.

[cijothomas]

It feels funny to not have DT per measurement, but technically, we can leave this to Aggregators that actually need it.

I cannot find a reason for knowing the timestamp at which the recording was made, as the output is an aggregation between a start and end time.
We'll need it in Exemplars(https://github.com/open-telemetry/opentelemetry-proto/blob/main/opentelemetry/proto/metrics/v1/metrics.proto#L528), and can be added if that feature is on, and as you said, it can be inside the individual aggregator itself.

[reyang]

LGTM.