This is the last step of the RFC to expand the Array.Parallel module, based on approved suggestion .

This PR includes:

• min
• minBy
• max
• maxBy
• sum
• sumBy
• average
• averageBy

And two functions on which the implementation relies on:

• reduce
• reduceBy (= map & reduce in a single step)

The benchmarks used for the choice of implementation are here https://github.com/T-Gro/Array.Parallel.Aggs.Benchmark/tree/main/Results .

Among others, the following dimensions were already evaluated and lead to the choice of existing implemented solution:

• Implement via reduce/reduceBy vs manual handcrafted impl for each
• Usage of ref-tuple vs. struct-tuple for minBy/maxBy operations
• Chunking the array vs. doing Parallel.For for each element of the array separately
• And most importantly, the amount of inlining:
  • Naturally, the fastest solution was to inline (and InlineIfLambda) everything
  • However, finally chosen solution inlines all the "one liners", but leaves the workhorse ("reduceBy") non-inlined. This was the second fastest option, very close to manually handrolled functions. And, will have smaller footprint on resulting codesize, as the compiler will only inline the one-liners and calls to "reduceBy", without repeating the body of it.