Computing the allocation statistics

[cshung]

In order to debug what is wrong with the POH allocation test case, I added this support in GCPerfSim to count how many bytes we allocated into the individual object heaps.

To start with the review, read the changes in GCPerfSim.cs.

The change accounts for the bytes allocated in individual object heaps in two ways. First, it is done in GetObjectSpec() and then it is done again in the Statistics class which is enabled only for debug, which allows me to be sure that the count is correct when I debug.

The change I did found a bug. In order for this accounting scheme to work, we need to make sure whenever GetObjectSpec() is called, there is exactly one allocation according to the spec happen downstream in JustMakeObject(). Just so it is clear, we need this constraint for another reason, if that is not true, then the allocation ratio arguments won't work.

The various thing that the original code did for ReferenceItemWithSize does not comply with this constraint, and therefore I changed the logic there, it will still create a linked list, but it will allocate exactly once.

The python code is driven by the fact that the GCPerfSim output is actually parsed by the infrastructure, so we need to make sure the infrastructure is aware of these extra outputs, and then py . lint do all the other changes.

[ivdiazsa]

Looks good! Just the typo I pointed and drop the linter's changes to bench_file.md. For some reason it messes with certain things like removing the example test_executables.

[Maoni0]

looks fine to me. a note for future reference, if you could refrain from having a significant chunk of your changes be formatting changes only that'd be much appreciated. I'm all for have formatting changes to make the formatting more consistent but having them mixed with actual functional changes makes CRs more work than it should be. I'm not gonna ask you to separate them out for this PR but please keep it in mind for future PRs. I'm guessing these changes simply come from whatever editor you are using that does the formatting. you can just do that once and submit a PR to cover all formatting changes.

I see this is marked as Draft, please let me know when you think this is ready for checkin.

[cshung]

Once the formatting change PR is merged, changes to GCPerfSim.cs should be purely functional changes.

[cshung]

This PR is subsumed by #1704