getMany performance tests

[pgte]

Preparing benchmarks to then try solving #163

[pgte]

My performance results for .getMany():

10 nodes, 10 blocks, 5 iterations
started
all nodes connected to each other
..........++++++++++..........++++++++++..........++++++++++..........++++++++++..........++++++++++
stopping
stopped
stats:
---------
mean: 3601.06
median: 3827.5
variance: 3656466.8163999994
standard deviation: 1912.189011682684
85th percentile: 5596

[pgte]

Click to view flamegraph (Chrome only):

[pgte]

A hot part is on the keypair dependency, which is used by libp2p-crypto:

→ npm ls keypair
[email protected] /Users/pedroteixeira/projects/ipfs/js-ipfs-bitswap
└─┬ [email protected]
  └─┬ [email protected]
    └── [email protected]

[pgte]

By following the thread of the keypair hint, above, I found out that the DHT is making heavy use of the unmarshalling of the public key to validate records.
A significant performance boost can be found if I turn the DHT off.

For comparison, with DHT:

------------------------
10 nodes, 10 blocks, 5 iterations
started
all nodes connected to each other
..........++++++++++..........++++++++++..........++++++++++..........++++++++++..........++++++++++
stopping
stopped
stats:
---------
mean: 3605.32
median: 3811.5
variance: 4057299.777599999
standard deviation: 2014.2740075769234
85th percentile: 5777

without DHT:

------------------------
10 nodes, 10 blocks, 5 iterations
started
all nodes connected to each other
..........++++++++++..........++++++++++..........++++++++++..........++++++++++..........++++++++++
stopping
stopped
stats:
---------
mean: 874.7
median: 963.5
variance: 93265.29000000004
standard deviation: 305.39366398142585
85th percentile: 1091

It signifies a x6 improvement without DHT..

[pgte]

This PR adds some benchmarking infra and a getMany benchmark test.
We can continue the improvements on a different PR.