feat: merge waits and recvs to wrCounter

[SoulPancake]

@rueian Please take a look at this
solves #807

[SoulPancake]

taking a look at the test failure, race

[rueian]

It looks like you didn't cast some of the return values correctly. I think we better have helper functions instead of scattering uint32( .... & 0xFFFFFFFF) and uint32( ... >>32) around the code base.

Maybe something like func IncrWaits() uint32, func DecrWaits() uint32, func DecrWaitsAndRecvs() (uint32, uint32), func LoadRecvs() uint32, and func LoadWaits() uint32

[SoulPancake]

I did that now, still miss what went wrong, can you help me out? @rueian

[rueian]

I did that now, still miss what went wrong, can you help me out? @rueian

^uint64(1) for decrement doesn't seem correct. Also, we need to merge p.DecrWaits() and p.IncrRecvs() into one p.DecrWaitsAndRecvs. That's the whole point of this improvement. Another thing is, my bad, we should make these helper functions private.

[rueian]

So we have two cases for decrements

1. p.decrWaits only decrements the lower 32bit. The upper 32bits should not be touched.
2. p.decrWaitsAndRecvs decrements both upper 32bit and lower 32bit in one operation.

[SoulPancake]

please take a look at it once @rueian

[SoulPancake]

Please re-review @rueian Sorry for the delay

[rueian]

Thanks so much @SoulPancake