feat(event pool): simplify realization of event pool

[ythosa]

Description

I have managed to improve the performance of the pool and improve its readability

Benchmarks

Old

BenchmarkEventPoolOneGoroutine-8        58460538                19.96 ns/op            0 B/op          0 allocs/op
BenchmarkEventPoolManyGoroutines-8           100          10746244 ns/op            4986 B/op         56 allocs/op
BenchmarkEventPoolSlowestPath-8              152           8189235 ns/op           27087 B/op       1034 allocs/op

New

BenchmarkEventPoolOneGoroutine-8        39543708                32.61 ns/op            0 B/op          0 allocs/op
BenchmarkEventPoolManyGoroutines-8           697           1724738 ns/op             876 B/op         15 allocs/op
BenchmarkEventPoolSlowestPath-8             1026           1038654 ns/op           24619 B/op       1007 allocs/op

[ythosa]

@kirillov6 @goshansmails please review!

[ythosa]

btw, i've tested realisation based on sync.Pool, it was too much faster than my new realisation, but the possibility of event dump was lost

[vadimalekseev]

@ythosa, hello!

You can use BenchmarkLightJsonReadPar to check how the performance has changed. On mac m1 this bench showed following result: old:

BenchmarkLightJsonReadPar-8            1        2256751708 ns/op         643.06 MB/s     4825008 B/op      48384 allocs/op

new:

BenchmarkLightJsonReadPar-8            1        4433031459 ns/op         327.37 MB/s     4272744 B/op      42659 allocs/op

Also, we can't use sync.Pool because insaneJSON uses own global pool. So, we need to refactor insaneJSON first, then we can test sync.Pool

[ythosa]

Wow, I haven't seen this benchmark. You are very cool...