fix: replace chunk locker with mutext

[notanatol]

Checklist

• I have read the coding guide.

Description

replace chunk locker mechanism with an stdlib's mutex

Related: #4341

[ldeffenb]

Are you planning any performance impact tests on this change? Seems to me that changing from a (presumably) chunk-specific lock to a global local on the syncChunkStore is going to seriously throttle (slow down) parallel operations in the node. But then again, something needs to better protect against the random issues we see in the relational integrity of the chunkStore.

[notanatol]

Yes, according to these two benchmarks (that both use wrapSync) the performance is comparable, most likely to the previous implementation of locker not doing any meaningful guarding.
current branch:

BenchmarkReservePutter-8  8481	    152460 ns/op	   27886 B/op	     396 allocs/op
BenchmarkCachePutter-8  6631	    180415 ns/op	   20683 B/op	     320 allocs/op

master:

BenchmarkReservePutter-8  7214	    154002 ns/op	   27914 B/op	     397 allocs/op
BenchmarkCachePutter-8   6877	    184118 ns/op	   20587 B/op	     322 allocs/op

however this task has been de-prioritized and it will be continued sometime next year, I'll be moving it into draft state now.

[istae]

this is being handled in this PR #4561