fix(core): stop rounding to nearest int in hrTimeTo*seconds() functions

[aabmass]

Which problem is this PR solving?

Fixes #3986

Short description of the changes

Remove Math.round() calls from hrTimeTo{Milli,Micro}seconds() functions in order to preserve higher precision from floating point parts.

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

• Existing tests are all passing
• Would like to test this against contrib repo but not sure how to do that

Checklist:

• Followed the style guidelines of this project
• Unit tests have been added
• Documentation has been updated

[codecov]

Codecov Report

Merging #4014 (0b4aaf9) into main (2b20565) will not change coverage.
The diff coverage is 100.00%.

❗ Current head 0b4aaf9 differs from pull request most recent head b356aa1. Consider uploading reports for the commit b356aa1 to get more accurate results

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4014   +/-   ##
=======================================
  Coverage   92.31%   92.31%           
=======================================
  Files         321      321           
  Lines        9189     9189           
  Branches     1953     1953           
=======================================
  Hits         8483     8483           
  Misses        706      706           

Files Changed	Coverage Δ
packages/opentelemetry-core/src/common/time.ts	97.18% <100.00%> (ø)

[dyladan]

Seems like it doesn't break any tests. Have you been able to try this fork in a real world app?

[aabmass]

Seems like it doesn't break any tests. Have you been able to try this fork in a real world app?

I tried it in a little test express server with the http instrumentation and it fixed the striping I mentioned in the issue. Is there a way to run the contrib repo tests against this branch?

[dyladan]

Seems like it doesn't break any tests. Have you been able to try this fork in a real world app?

I tried it in a little test express server with the http instrumentation and it fixed the striping I mentioned in the issue. Is there a way to run the contrib repo tests against this branch?

Not an easy way that I can think of. We could release a -alpha tag version and push it to NPM.

I am very familiar with the trace SDK and don't expect this to be a problem, but can someone with more familiarity with the metrics SDK sanity check this? @legendecas @pichlermarc

[legendecas]

The only call site of the conversion functions is merging two accumulations, e.g. https://github.com/open-telemetry/opentelemetry-js/blob/main/packages/sdk-metrics/src/aggregator/LastValue.ts#L70. I believe it is fine to accept non-integer numbers here.

[aabmass]

Test failures api-eol-node-test are unrelated to this PR, they are failing on HEAD commit 2b20565

Unexpected token {
npm ERR! Test failed.  See above for more details.

[legendecas]

I'm going to land this and track the api-eol failure in #4029.