Fix division and limit benchmark-based usage factors

[mbenke]

Sometimes usage benchmark results can get misreported. This PR introduces sanity checks for usage factors based on these benchmark results, limiting them to the [0.1,2] range.

This PR also modifies initial usage factor calculation to take into account the fact that with usage benchmarks, (in contrast to efficiency benchmarks), faster machines correspond to smaller values.

Additionally, we prevent dividing by 0 (or other very small value of the usage benchmark).

Note: this PR is basically #4803 rebased to develop after merging #4802

[shadeofblue]

[codecov]

Codecov Report

Merging #4827 into develop will increase coverage by 0.04%.
The diff coverage is 100%.

@@             Coverage Diff             @@
##           develop    #4827      +/-   ##
===========================================
+ Coverage     89.2%   89.24%   +0.04%     
===========================================
  Files          226      226              
  Lines        20775    20567     -208     
===========================================
- Hits         18532    18355     -177     
+ Misses        2243     2212      -31