perf: protocol is very slow comapred to other implementations

[vladopajic]

perf takes too much time (it gets time-out) while running in test-plans repo (libp2p/test-plans#658, working pr: libp2p/test-plans#659)

---

if 900719925 value is used for download bytes, following performance is measured:

tcp + yamux + noise:
go server  + go client  = 0.8s
go server  + nim client = 14s
nim server + go client  = 11.5s
nim server + nim client = 10-14s

tcp + mplex + noise
nim server + nim client = 14s

[vladopajic]

prs addressing this issue:

#1488
#1491
#1492
#1497

---

with these prs the performance is improved by approximately 46.5%.

below is table with total duration needed to complete perf (tcp + yamux + noise) with same load in 5 measuring samples:

before: 18.6s, 17.9s, 18.6s, 17.5s, 18.5s
after#1:   8.8s, 10.2s,  9.6s, 10.1s, 10.0s

avrg(before):  18.22s
avrg(after#1):  9.74s

[arnetheduck]

enable -d:lto and --stacktrace:off (or switch to https://github.com/status-im/nim-libbacktrace) for a free performance boost.

[vladopajic]

enable -d:lto and --stacktrace:off (or switch to https://github.com/status-im/nim-libbacktrace) for a free performance boost.

tested --stacktrace:off it does not have any difference;
and these measurments already used lto

compile command is here and that is also the binary used to run perf benchmarking.

[vladopajic]

with: #1510

after#2:  8.1s 8.2s 8.3s 8.3s 8.0s

avrg(after#2):  8.18s

current performance improvement is approximately 55.13%.