[sdk-metrics] Fix a race condition when a metric is deactivated and re-activated in the same collection cycle

[CodeBlanch]

Fixes #5907

Changes

• Use TryUpdate instead of TryRemove when processing a removed metric so that the lookup dictionary is only modified if a new metric hasn't already been enrolled

Merge requirement checklist

• CONTRIBUTING guidelines followed (license requirements, nullable enabled, static analysis, etc.)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 86.41%. Comparing base (6250307) to head (201943f).
Report is 357 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5908      +/-   ##
==========================================
+ Coverage   83.38%   86.41%   +3.03%     
==========================================
  Files         297      258      -39     
  Lines       12531    11345    -1186     
==========================================
- Hits        10449     9804     -645     
+ Misses       2082     1541     -541     

Flag	Coverage Δ
unittests	?
unittests-Project-Experimental	86.39% <100.00%> (?)
unittests-Project-Stable	86.35% <100.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...rc/OpenTelemetry/Metrics/Reader/MetricReaderExt.cs	92.24% <100.00%> (ø)

... and 236 files with indirect coverage changes

[Kielek]

LGTM.

Do you think that it is worth to document in CHANGE:LOG?

[CodeBlanch]

@Kielek I don't think we need to mention anything in CHANGELOG because there really isn't any impact to users if they run into the race. Metrics will continue to work even if the wrong lookup entry is removed.