Atomic async writes

[mcollina]

Fixes: #96

[mcollina]

I resolved most of the preformance regressions introduced for the async and sync case. However it still degrades the performance if minLength is set. I think this is good to land anyway as it improves things for the basic case!

This branch:

benchSonic*1000: 455.379ms 
benchSonicSync*1000: 6.695s
benchSonic4k*1000: 450.397ms                                                                                                                                                                   
benchSonicSync4k*1000: 330.373ms
benchCore*1000: 4.554s
benchConsole*1000: 9.565s
benchSonic*1000: 378.047ms
benchSonicSync*1000: 6.684s
benchSonic4k*1000: 440.553ms
benchSonicSync4k*1000: 325.768ms
benchCore*1000: 4.518s
benchConsole*1000: 9.638s

master:

benchSonic*1000: 486.568ms
benchSonicSync*1000: 6.806s
benchSonic4k*1000: 437.865ms
benchSonicSync4k*1000: 296.988ms
benchCore*1000: 4.506s
benchConsole*1000: 9.595s
benchSonic*1000: 404.316ms
benchSonicSync*1000: 6.756s
benchSonic4k*1000: 419.896ms
benchSonicSync4k*1000: 293.911ms
benchCore*1000: 4.526s
benchConsole*1000: 9.163s

Those were taken on Node.js v16.

[ronag]

Typo?

[ronag]

@mcollina pino still suffers from this problem since it has sonic-boom v1 as dependency. Could we bump it on pino v6?

[mcollina]

You can use sonic-boom with pino@6, just instantiate one and pass it as a stream.

Unfortunately there are a few semver-major changes in there.