Optimize memchr with NEON

[ywxt]

Result:

1. memchr:

Benchmark	Length	Before (ns/iter)	After (ns/iter)	Speedup (x)
fwd_match_first	16	3.80	3.45	1.10
	32	3.79	3.45	1.10
	64	3.79	3.80	1.0
	256	3.80	3.80	1.0
	1k	3.80	3.80	1.0
	4k	3.79	3.80	1.0
	64k	3.79	3.80	1.0
fwd_match_last	16	9.66	9.32	1.04
	32	10.01	9.66	1.04
	64	12.65	11.41	1.11
	256	28.74	14.15	2.03
	1k	89.09	31.60	2.82
	4k	336.51	90.49	3.72
	64k	5098.57	1761.04	2.90
fwd_match_none	16	3.45	3.11	1.11
	32	4.17	3.45	1.21
	64	6.25	3.45	1.81
	256	21.11	6.56	3.22
	1k	81.11	27.22	2.98
	4k	321.08	92.32	3.48
	64k	5107.51	1389.31	3.68
rev_match_first	16	10.39	8.06	1.29
	32	12.42	8.51	1.46
	64	16.67	10.92	1.53
	256	30.79	13.06	2.36
	1k	93.71	31.24	3.00
	4k	350.16	85.71	4.09
	64k	5364.25	1638.84	3.27
rev_match_last	16	4.36	2.11	2.07
	32	4.35	2.11	2.06
	64	4.30	3.14	1.37
	256	4.34	3.12	1.39
	1k	4.29	3.13	1.37
	4k	4.30	3.12	1.38
	64k	4.29	3.34	1.28
rev_match_none	16	4.23	2.21	1.91
	32	6.79	2.25	3.02
	64	7.78	2.87	2.71
	256	23.83	6.99	3.41
	1k	86.16	23.19	3.72
	4k	346.42	83.58	4.14
	64k	5368.44	1704.41	3.15

2 pattern

Benchmark	Before (ns/iter)	After (ns/iter)	Speedup (x)
find_1byte_str_long_match_end	5180.74	1182.60	4.38
rfind_char_long_nomatch	5351.41	1279.11	4.18
rfind_1byte_str_long_nomatch	5391.75	1302.14	4.14
find_char_long_match_end	5096.61	1242.24	4.10
find_1byte_str_long_nomatch	5172.59	1286.60	4.02
find_char_long_nomatch	5104.69	1318.24	3.87
split_char_sparse	14326.42	4654.07	3.08
split_1byte_str_sparse	16889.26	7111.13	2.37
split_char_multibyte_haystack	69253.86	62913.94	1.10
split_1byte_str_multibyte_haystack	80609.83	78901.17	1.02
find_str_worst_case	1482.15	1460.42	1.01
split_char_dense	42748.78	43158.21	0.99
find_str	7090.74	7093.10	1.00
find_char_early_return	7787.22	7784.51	1.00
starts_with_str	4597.53	4599.39	1.00
starts_with_char	4598.04	4598.76	1.00
ends_with_str	4598.14	4599.04	1.00
ends_with_char	4597.97	4600.08	1.00
find_1byte_str_short_haystack	9242.62	9314.57	0.99
split_1byte_str_dense	54059.19	55237.39	0.98
find_1byte_str_early_return	11512.55	12162.06	0.95
find_char_short_haystack	5313.39	5569.64	0.95
rfind_str_worst_case	23512.81	24622.20	0.95
rfind_str	5708.62	6225.30	0.92

[rustbot]

r? @LawnGnome

rustbot has assigned @LawnGnome.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: libs
• libs expanded to 12 candidates
• Random selection from JohnTitor, LawnGnome, Mark-Simulacrum, clarfonthey, nia-e

[ywxt]

r? @Amanieu

[Amanieu]

I don't think that doing something neon-specific here is the right approach. Instead we should consider importing the implementation from the memchr crate directly into the standard library.

See the discussion in #159090.

cc @BurntSushi @joboet