Previously, we had some white listing lists for allocations that were assigned
into globals during a benchmarks running. Now we instead run two different
iterations with the second running an additional time. Then we subtract the
counts.
This enables me to get rid of the whitelist.
Make sure all Python code in the repo specifies which exceptions to
catch when using `except:`.
Regressions can be catched using `flake8-blind-except` going forward.
* E101: indentation contains mixed spaces and tabs
* E111: indentation is not a multiple of four
* E128: continuation line under-indented for visual indent
* E302: expected 2 blank lines, found 1
* W191: indentation contains tabs