even faster unlock in contention

[BusyJay]

This is an alternative more aggressive implementation of idea #461.

Compared to #461, this PR

• maintains parked bit on waiter side, so that waker doesn't have to atomic operation twice.
• reset all lock states back to 0 when unlock. This makes fast lock more likely succeed during high contention.
• set PARKED_BIT even waiter is prevented from sleep, so that more threads can be woken up during contention to compete for progress.

Running cargo run --bin mutex --release -- 9:36:9 5 5 2 2:

Running with 9 threads

name	average	median	std.dev.
parking_lot::Mutex (this pr)	472.082 kHz	475.676 kHz	13.820 kHz
parking_lot::Mutex (pr 461)	405.134 kHz	406.469 kHz	9.105 kHz
parking_lot::Mutex (master)	364.841 kHz	365.127 kHz	11.269 kHz
std::sync::Mutex	769.754 kHz	767.714 kHz	23.908 kHz
pthread_mutex_t	982.966 kHz	989.991 kHz	31.816 kHz

Running with 18 threads

name	average	median	std.dev.
parking_lot::Mutex (this pr)	352.643 kHz	353.062 kHz	7.842 kHz
parking_lot::Mutex (pr 461)	268.530 kHz	268.355 kHz	5.586 kHz
parking_lot::Mutex (master)	82.786 kHz	82.975 kHz	2.435 kHz
std::sync::Mutex	389.199 kHz	394.549 kHz	21.358 kHz
pthread_mutex_t	482.005 kHz	489.225 kHz	26.404 kHz

Running with 27 threads

name	average	median	std.dev.
parking_lot::Mutex (this pr)	231.691 kHz	231.891 kHz	3.014 kHz
parking_lot::Mutex (pr 461)	185.802 kHz	186.233 kHz	2.598 kHz
parking_lot::Mutex (master)	28.246 kHz	28.306 kHz	0.443 kHz
std::sync::Mutex	280.553 kHz	280.014 kHz	10.115 kHz
pthread_mutex_t	311.815 kHz	311.582 kHz	8.409 kHz

Running with 36 threads

name	average	median	std.dev.
parking_lot::Mutex (this pr)	157.511 kHz	157.183 kHz	1.764 kHz
parking_lot::Mutex (pr 461)	134.010 kHz	133.784 kHz	1.509 kHz
parking_lot::Mutex (master)	22.055 kHz	22.059 kHz	0.150 kHz
std::sync::Mutex	193.672 kHz	195.078 kHz	10.369 kHz
pthread_mutex_t	224.436 kHz	225.767 kHz	12.334 kHz

Running cargo run --bin rwlock --release -- 36 9 5 5 2 2

name	write	read
parking_lot::RwLock (this pr)	6805.309 kHz	1334.018 kHz
parking_lot::RwLock (pr 461)	6121.347 kHz	968.373 kHz
parking_lot::RwLock (master)	628.062 kHz	954.938 kHz
seqlock::SeqLock	648.979 kHz	152225.000 kHz
pthread_rwlock_t	1678.253 kHz	376.558 kHz

Using lock-bench, cargo run --release 32 2 10000 100:

name	avg	min	max
std::sync::Mutex	30.795793ms	28.369313ms	33.668656ms
parking_lot::Mutex (this pr)	32.460918ms	29.497656ms	34.424228ms
parking_lot::Mutex (pr 461)	40.800542ms	37.16543ms	44.621677ms
parking_lot::Mutex (master)	206.836045ms	183.902676ms	213.697023ms
spin::Mutex	63.898884ms	58.45244ms	74.323676ms
AmdSpinlock	70.131547ms	65.356139ms	83.456119ms

name	avg	min	max
std::sync::Mutex	30.52266ms	28.69828ms	34.945486ms
parking_lot::Mutex (this pr)	31.257888ms	29.648337ms	33.955286ms
parking_lot::Mutex (pr 461)	41.146074ms	38.453175ms	42.433051ms
parking_lot::Mutex (master)	210.387478ms	187.38791ms	215.752182ms
spin::Mutex	62.823716ms	54.801191ms	74.31628ms
AmdSpinlock	68.937325ms	55.406785ms	80.83359ms