feat(Benchmarks): Add real world dup blocks test

[hannahhoward]

Goals

Build a benchmark that can replicate the problems that #8 has with LS-ing Wikipedia

Implementation

• add a delay generator that similates real world latencies one might encounter on the internet
• modify virtual network to accept different latencies for different
  peers based on using NextWaitTime on passed delay
• modify dup_blocks_test subtestDistributeAndFetch to accept a custom
  delay
• Add a real world benchmarks that simulates the kinds of problems one might
  encounter bitswaping with a long lived session and a large swarm of
  peers with real world latency distributions (that causes [WIP] Work towards improving the duplicate blocks issue #8 not to
  function well in practice)

For Discussion

The goal here is to have a good benchmark to compare against as we continue to develop improvements to bitswap.

One thing not addressed here are transfer speeds (though you can think of latency in this test as really being time to first byte -- i.e. traditional latency -- + download time). Is that worth addressing seperately.

Benchmarks with results posted in comment.

child of ipfs/kubo#5723

[hannahhoward]

Benchmark Results master vs feat/reduce-dupes

These are comparisons of each of the three network conditions on both master and Jeremy's branch.

Super great network

Everyone’s coming in around 60ms latency

Branch	Duplicate Blocks	Total Time
master	52804	29.317s
feat/reduce-dupes	1192	7.563s

Jeremy PR kicks butt cause it makes less requests / doesn't
deal have to deal with dupes

So So Network

40% at 60ms latency, 30% at 200ms latency, 30% at 800ms latency

Branch	Duplicate Blocks	Total Time
master	30704	16.599s
feat/reduce-dupes	1192	33.628s

Master gets faster because it doesn't deal with as many dupes
Meanwhile Jeremy PR still no extra dupes but now quite a bit slower than
master

Worst case scenario (a.l.a. Wikipedia)

40% at 60ms latency, 30% at 200ms latency, 30% at 4000ms latency

Branch	Duplicate Blocks	Total Time
master	38885	7.066s
feat/reduce-dupes	20045	72.017s

Now Jeremy's PR is getting comparable dupes due to broadcasts.
Moreover, it's now almost 10x slower than master

[hannahhoward]

I would be curious on @whyrusleeping 's feedback if this seems like a good test.

[hannahhoward]

Seeking 1 LGTM from anyone to merge, since this is just benchmarks

[michaelavila]

LGTM