[testing] Lower default r2r NativeHashtable fill factor to improve runtime lookup performance

[kg]

In some startup measurements we spend a lot of time looking up r2r entry points. One potential remedy for this is to lower the fill factor for the hashtable used to perform these lookups, so that each bucket contains fewer methods and we have fewer methods to check as a result. This will raise the size of r2r images by Some Amount, and it's possible that changing it could break something, so I'm opening this draft PR to do an outerloop run and see if anything falls over.

[kg]

/azp run runtime-coreclr crossgen2 outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[kg]

Looks like this change fails the determinism checks (the generated R2R binaries change in contents but not size), but nothing else breaks, which is a good sign. We can do this if we need to.

My local investigations indicate that we don't actually need to do this at present so I'm closing it now that I have the data from the CI run.