Use jetscii for SIMD searching.

[kivikakk]

This results in a very moderate (1.01x) slow-down on aarch64, but a slightly less moderate (1.05x) speed-up on x86_64 — 57% of our HTML formatting time in the benchmark is spent in html::escape.

With shepmaster/jetscii#65 the aarch64 slow-down changes to a 1.05x speed-up, but most real workloads are probably still on x86_64, so I don't mind that we're not there yet. I'd love us to be, though!

[github-actions]

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
./bench.sh ./comrak-f4ca423	111.2 ± 1.1	109.6	113.7	1.92 ± 0.05
./bench.sh ./comrak-main	116.1 ± 1.5	113.7	119.3	2.00 ± 0.06
./bench.sh ./pulldown-cmark	57.9 ± 1.4	56.7	65.8	1.00
./bench.sh ./cmark-gfm	86.8 ± 7.3	81.2	105.4	1.50 ± 0.13
./bench.sh ./markdown-it	258.9 ± 1.7	255.6	262.5	4.47 ± 0.11

Run on Sun Oct 19 03:38:29 UTC 2025