Improve StepBucketHistogram

[lenin-jaganathan]

This polishes/fixes gh-3749

• eliminate null check on noValue
• to support exporting cumulative buckets, more so in a view that this can be leveraged in other libraries as well. (see - DeltaHistogram in SignalFx registry doesn't align with count and total #3774)
• fix explicit histogram buckets (SLO), was zero previously

[shakuzen]

Does this need to be protected? It looks like it could be package-private instead.

[lenin-jaganathan]

StepBucketHistogram is part of the distribution package so this needs to be protected to use this constructor.

[shakuzen]

It looks like OTelCollectorIntegrationTest is failing. Perhaps something has changed in a recently published version of the OTel Collector docker image.

[lenin-jaganathan]

@shakuzen The failures in the OTEL collector test are mostly related to your comment here (at least I think so).

Basically, the test expects the timer to have a name like "test_timer_count", but it seems the name is changed to test_timer__count. The same goes for the counter as well, where test_counter is now test_counter_total. Haven't dug in deep but I don't see any related PR's there.

[lenin-jaganathan]

@shakuzen / @jonatan-ivanov Does it makes sense for the OtlpStepTimer to be re-named as OtlpDeltaTimer? This is something that I am not able to choose either but I have a slight preference for renaming this as OtlpStepMeters as OtlpDeltaMeter

[jonatan-ivanov]

@lenin-jaganathan Thank you very much for the PR!
I fixed the build on all of our branches, also rebased this PR to main, see details here: #3796

The summary is: our tests depend on the (wrong) behavior of the OTel collector, some of them were fixed which was a breaking change, so our tests were also broken.

[jonatan-ivanov]

Does it makes sense for the OtlpStepTimer to be re-named as OtlpDeltaTimer? This is something that I am not able to choose either but I have a slight preference for renaming this as OtlpStepMeters as OtlpDeltaMeter

Sorry, I missed this comment while I was reviewing/played with this.
My two cents:

• Since it is package-private, it does not have a big significance, we can change anytime.
• To me the Step* makes a little bit more sense from the perspective of consistency. I'm thinking that instead of renaming, maybe it would make sense calling out the reason why it does not extend StepTimer in the Javadoc .