fix(sdk-metrics): allow single bucket histograms

[pichlermarc]

Which problem is this PR solving?

When creating a histogram with an empty bucket via the API, the SDK throws. Such a histogram can be useful if just min/max/sum/count are needed. Since previously, the only way to set up an aggregation would have been via Views, it was fine for ExplicitBucketHistogramAggregation to throw (View creation happens on SDK setup). Now that the same thing can happen via the API we need to ensure that we do not throw.

Further, reading through the spec it looks like single-bucket histograms are actually permitted. Therefore, I didn't change it to log a warning instead.

I added a few more tests to ensure that this special case is covered works with the the existing aggregators.

Type of change

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

How Has This Been Tested?

• Added unit tests

[codecov]

Codecov Report

Merging #4456 (5d9da2b) into main (f6712fd) will decrease coverage by 0.20%.
The diff coverage is 100.00%.

❗ Current head 5d9da2b differs from pull request most recent head ce239b6. Consider uploading reports for the commit ce239b6 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4456      +/-   ##
==========================================
- Coverage   93.12%   92.93%   -0.20%     
==========================================
  Files          82      259     +177     
  Lines        1658     6991    +5333     
  Branches      341     1459    +1118     
==========================================
+ Hits         1544     6497    +4953     
- Misses        114      494     +380     

Files	Coverage Δ
packages/sdk-metrics/src/view/Aggregation.ts	100.00% <100.00%> (ø)

... and 178 files with indirect coverage changes