Benchmark to ensure local times are equivalent

[pbiggar]

So far some optimization has happened, resulting in fixing bugs that reduced runtime for fizzbuzz by 100x.

That said, it seems like for basic benchmarks, Dark in F# is 9x slower than Dark in OCaml. This isn't a fair comparison since Dark in F# is async and Dark in OCaml isn't. But we still need Dark in F# to be operationally efficient.

Remaining things to check:

• load test test fizzbuzz, see if it actually is better or worse, load-wise
• benchmark and loadtest things that use httpclient
• benchmark and loadtest things that use json
• benchmark and loadtest things that use the DB
• run a profiler over the fizzbuzz test, see if there are any low-hanging fruit

[pbiggar]

Closing in favor of #3624