Logs: Scopes, State, & FormattedMessage follow-up

[CodeBlanch]

Changes

Code review, unit tests, and performance tweaks from #1869.

I did some benchmarking and it turns out the dynamic method was actually slower & using more memory than the boxing we had before, so I removed it.

With parsing:

Method	NumberOfMessages	Mean	Error	StdDev	Gen 0	Gen 1	Gen 2	Allocated
Parsing	100	40.38 us	0.753 us	0.629 us	4.0894	-	-	33.52 KB

With boxing:

Method	NumberOfMessages	Mean	Error	StdDev	Gen 0	Gen 1	Gen 2	Allocated
Boxing	100	34.84 us	0.403 us	0.336 us	3.1128	-	-	25.7 KB

Looking at the struct used for the log extensions (LogInformation, LogWarning, etc.), the results make sense. The box only has to move the three pointers. The KVP data is already in the object[] on the heap. The parser had to allocate a new list, and then loop over the KVPs to add them one-by-one. The mean time gets worse for each item added to the state. Oh well, they all can't be zingers.

TODOs:

• CHANGELOG.md updated for non-trivial changes

[codecov]

Codecov Report

Merging #1883 (40e2e4e) into main (41a3371) will increase coverage by 0.85%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #1883      +/-   ##
==========================================
+ Coverage   83.34%   84.19%   +0.85%     
==========================================
  Files         188      188              
  Lines        6141     6092      -49     
==========================================
+ Hits         5118     5129      +11     
+ Misses       1023      963      -60     

Impacted Files	Coverage Δ
src/OpenTelemetry/Logs/LogRecord.cs	92.50% <100.00%> (+10.00%)	⬆️
src/OpenTelemetry/Logs/OpenTelemetryLogger.cs	93.93% <100.00%> (+67.11%)	⬆️
...c/OpenTelemetry/Logs/OpenTelemetryLoggerOptions.cs	87.50% <100.00%> (ø)
...emetry.Api/Internal/OpenTelemetryApiEventSource.cs	76.47% <0.00%> (-5.89%)	⬇️

[reyang]

Ping back from #1325 and #1328.