Potential compile time regression in item-types checking

[Mark-Simulacrum]

Diff: 1cdc0fb...144af3e

Perf.rlo graph, though it basically just shows that item-types checking regressed from 0.67 to 1.006.

cc @brson (the commits in the range are from #36807).

IRC discussion has led @brson to believe that this is a result of memchr changes.

[brson]

The only thing in that patch that looks remotely performance sensitive to me is the inlining around memchr. If I accidentally switched the memchr implementation to the slow version, that would also explain it.

[brson]

The regression disappeared.

[Mark-Simulacrum]

I guess this can then be closed, probably just noise.