test(benchmark): Replace matcha with nanobench

[jhermsmeier]

This replaces matcha with nanobench, which will enable us to easily compare between changes and versions with nanobench-compare.

$ node benchmark/bencode.js
NANOBENCH version 2
> node benchmark/bencode.js

# bencode.encode() [buffer] ⨉ 10000
ok ~474 ms (0 s + 473525130 ns)

# bencode.encode() [utf8] ⨉ 10000
ok ~8.74 s (8 s + 743247355 ns)

# bencode.encode() [ascii] ⨉ 10000
ok ~699 ms (0 s + 699163950 ns)

# bencode.encode() [binary] ⨉ 10000
ok ~2.32 s (2 s + 322715866 ns)

# bencode.decode() [buffer] ⨉ 10000
ok ~199 ms (0 s + 198994278 ns)

# bencode.decode() [utf8] ⨉ 10000
ok ~6.01 s (6 s + 5084972 ns)

# bencode.decode() [ascii] ⨉ 10000
ok ~411 ms (0 s + 411197073 ns)

# bencode.decode() [binary] ⨉ 10000
ok ~137 ms (0 s + 137260635 ns)

all benchmarks completed
ok ~19 s (18 s + 991189259 ns)

One thing to note is also that this gives us higher resolution timing on an iteration level; for example the ~8.74 s for the UTF8 encoding run might seem terribly slow, but that's only 0.874 ms per iteration – on a 27 KB torrent file.