Add the Exponential Histogram Aggregator.

[MadVikingGod]

This is the final part of #2966.

This adds the Aggregator and the mechanism to configure it.

Some notes when reading this:

• When Scale is 20, every normal floating point can be mapped neatly into a bucket at some offset.
• When Scale is -10 every number maps to bucket -1 or 0.
• The spec says we can round subnormal numbers to the smallest normal number for bucketing. This is done because most of the math.Log returns inaccurate results for subnormal numbers.

[codecov]

Codecov Report

Merging #4245 (e626be7) into main (f67ecb3) will increase coverage by 0.1%.
The diff coverage is 92.5%.

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #4245     +/-   ##
=======================================
+ Coverage   83.0%   83.2%   +0.1%     
=======================================
  Files        217     218      +1     
  Lines      17166   17460    +294     
=======================================
+ Hits       14264   14538    +274     
- Misses      2611    2629     +18     
- Partials     291     293      +2     

Files Changed	Coverage Δ
sdk/metric/pipeline.go	88.3% <10.0%> (-2.4%)	⬇️
sdk/metric/aggregation/aggregation.go	80.9% <72.7%> (-3.0%)	⬇️
...metric/internal/aggregate/exponential_histogram.go	96.2% <96.2%> (ø)
sdk/metric/internal/aggregate/aggregate.go	100.0% <100.0%> (ø)

... and 1 file with indirect coverage changes

[pellared]

Can we add a "test example" to demonstrate how one can configure an exponential histogram?

[MadVikingGod]

Can we add a "test example" to demonstrate how one can configure an exponential histogram?

Done: added in b41b1e4

[pellared]

Changelog is missing 😉

[MrAlias]

Can you add "inOut" to the ignored codespell words?