Improve like kernel by ~2%

[psvri]

Which issue does this PR close?

Closes #.

Rationale for this change

While experimenting with improving like kernel, I found that we can check for like pattern characters much more efficiently using memchr2.

What changes are included in this PR?

Reworks like kernel logic. Perf numbers in my system is as below

regexp                  time:   [14.796 ms 14.914 ms 15.039 ms]
                        change: [-3.3875% -2.2208% -1.0196%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 18 outliers among 100 measurements (18.00%)
  15 (15.00%) high mild
  3 (3.00%) high severe

regexp scalar           time:   [9.8261 ms 9.9221 ms 10.018 ms]
                        change: [-1.6210% -0.3701% +0.8929%] (p = 0.57 > 0.05)
                        No change in performance detected.

Are there any user-facing changes?

No

[tustvold]

Seems reasonable to me