Use Extrema type for Histogram min/max

[MrAlias]

Resolve #3201

Prevent two allocations for every HistogramDataPoint that is created by replacing the Min and Max pointer fields with a new Extrema type. Meaning, this

type HistogramDataPoint struct {
	Min *float64
	Max *float64
}

becomes

type HistogramDataPoint struct {
	Min Extrema
	Max Extrema
}

with Extrema declared as:

// Extrema are the minimum or maximum values of a data set.
type Extrema struct {
	// value is the extrema value.
	value float64

	// valid is true if Value has been assigned a value. It is false if Value
	// is the zero-value.
	valid bool
}

See this gist for a comparison of the performance among the approaches proposed in #3201.

[codecov]

Codecov Report

Merging #3550 (6436bb3) into main (c0fb8de) will increase coverage by 0.0%.
The diff coverage is 100.0%.

Additional details and impacted files

@@          Coverage Diff          @@
##            main   #3550   +/-   ##
=====================================
  Coverage   79.4%   79.4%           
=====================================
  Files        170     170           
  Lines      12628   12645   +17     
=====================================
+ Hits       10030   10047   +17     
  Misses      2388    2388           
  Partials     210     210           

Impacted Files	Coverage Δ
...s/otlp/otlpmetric/internal/transform/metricdata.go	100.0% <100.0%> (ø)
sdk/metric/internal/histogram.go	100.0% <100.0%> (ø)
sdk/metric/metricdata/data.go	57.1% <100.0%> (+57.1%)	⬆️
sdk/metric/metricdata/metricdatatest/assertion.go	84.1% <100.0%> (+0.4%)	⬆️
...dk/metric/metricdata/metricdatatest/comparisons.go	97.5% <100.0%> (+<0.1%)	⬆️
semconv/internal/v2/http.go	100.0% <100.0%> (ø)

[MrAlias]

We should add benchmarks here to show qualitative values of different proposals.

[MrAlias]

We should add benchmarks here to show qualitative values of different proposals.

https://gist.github.com/MrAlias/fefe010bbee8671e65d4540af44911f5