Added: Speed test for xxhash3_64

[Sewer56]

This PR adds a speed test for xxhash3, latest implementation from twox-hash, and renames the existing test to xxhash64.
Some results from my machine, AMD Ryzen 9 5900X, 32GB DDR4-3000 (16-17-17-35)

I feel like there should be some sort of disclaimer somewhere around the benchmark code that this measures the Hasher streaming interface, rather than pure throughput. The code is omitting the oneshot APIs, where certain algorithms like the xxh3 implementation benefit severely.

In any case, I did also benchmark xxh3 oneshot out of curiosity. xxh3 oneshot sits between wyhash and rapidhash for most inputs; before starting to run away from other stable algorithms at 512 bytes.

Short String Performance (2-64 characters)

Implementation	str_2	str_8	str_16	str_64
fxhash	1.57ns	1.36ns	1.77ns	3.83ns
gxhash*	2.09ns	2.08ns	2.08ns	2.54ns
rustc-hash	2.24ns	2.03ns	1.99ns	3.46ns
rapidhash_raw	2.67ns	2.47ns	2.47ns	4.08ns
rapidhash	2.75ns	2.44ns	2.49ns	4.09ns
wyhash_raw	3.15ns	3.19ns	3.29ns	4.06ns
ahash	3.59ns	3.59ns	3.82ns	5.55ns
wyhash	3.60ns	3.58ns	3.81ns	4.78ns
metrohash*	4.02ns	3.61ns	4.49ns	8.01ns
t1ha*	4.70ns	4.67ns	4.47ns	7.87ns
default	5.56ns	6.34ns	7.37ns	13.81ns
xxhash64	8.71ns	8.02ns	8.63ns	12.55ns
farmhash*	12.24ns	15.10ns	15.40ns	34.21ns
seahash	13.23ns	8.91ns	9.65ns	13.69ns
xxhash3_64	21.88ns	21.78ns	21.53ns	23.28ns
highwayhash	33.98ns	32.63ns	29.31ns	31.60ns

Long String Performance (100+ characters)

Implementation	str_100	str_177	str_256	str_1024	str_4096
rapidhash_raw	4.38ns	7.65ns	9.05ns	31.50ns	128.07ns
gxhash*	4.36ns	5.31ns	5.90ns	16.71ns	54.33ns
rapidhash	4.48ns	7.16ns	8.44ns	31.57ns	130.17ns
rustc-hash	4.67ns	7.42ns	9.62ns	39.64ns	175.03ns
fxhash	6.81ns	14.62ns	21.00ns	123.32ns	549.44ns
wyhash_raw	7.00ns	10.19ns	12.96ns	47.94ns	193.76ns
wyhash	7.34ns	11.47ns	13.90ns	52.58ns	204.68ns
t1ha*	9.50ns	12.20ns	15.92ns	51.76ns	206.03ns
metrohash*	10.87ns	16.75ns	18.10ns	54.97ns	202.29ns
ahash	12.71ns	16.25ns	20.44ns	87.60ns	357.70ns
xxhash64	14.31ns	19.91ns	22.28ns	66.61ns	235.66ns
seahash	16.92ns	25.39ns	31.93ns	114.53ns	427.51ns
default	18.54ns	31.04ns	40.10ns	150.11ns	577.65ns
xxhash3_64	27.80ns	33.27ns	28.49ns	48.89ns	96.06ns
highwayhash	39.86ns	44.03ns	41.78ns	84.37ns	251.55ns
farmhash*	45.33ns	58.66ns	61.79ns	103.12ns	275.50ns

Miscellaneous Types (u64 and object)

Implementation	u64	object
fxhash	281ps	7.43ns
rustc-hash	404ps	4.36ns
gxhash*	539ps	6.31ns
wyhash	819ps	14.40ns
rapidhash	999ps	10.90ns
metrohash*	1.06ns	37.65ns
wyhash_raw	3.40ns	-
seahash	8.53ns	68.82ns
xxhash64	8.58ns	40.99ns
default	8.71ns	35.28ns
farmhash*	10.09ns	101.06ns
highwayhash	34.85ns	52.22ns
xxhash3_64	21.24ns	46.27ns

[hoxxep]

Thanks for this! We could probably remove the base XxHash tests and simply replace them all with XxHash3_64 if they're deemed superior to the old XxHash?

The upgraded twox-hash package is breaking the MSRV tests. I'll move the benchmarking into another crate (or behind a bench feature?) when I add a few more hash comparisons for recent crates such as foldhash.