performance tests: define a baseline

I think we should define some performance requirements for bitswap (like, for example: "for 10 bitswaps exchanging 1000 blocks between themselves, they must execute this in X seconds or less"), and then apply these to the already existing `test/benchmarks`.

This is almost done (`test/benchmarks` already can be parametrised to spawn a network of x nodes exchanging y blocks), so I think this is just a matter of defining the baseline to:

a) prevent performance regressions
b) incorporate improvements

@VictorBjelkholm @vmx Thoughts on this?