Fix _customContainsEquatableElement implementation for Stride* types

[xwu]

Recently reported on the Swift forums, Stride* types fail to account for the scenario of striding "backwards" in its implementations of _customContainsEquatableElement when determining if an element is contained within the given interval.

This PR fixes that issue. Specifically, we modify the implementation to branch on the sign of _stride. Recall that we deliberately allow striding in the "opposite" direction of that specified by from and to/through, which results in an empty sequence—in that case, this PR results in _customContainsEquatableElement correctly returning false regardless of element without special cases.

Resolves SR-15384.

[xwu]

@swift-ci test

[stephentyrone]

I'm slightly confused as to why these functions only return either false or nil. It appears that this makes us always fall back on the predicate version of contains when the element is present; did we just punt on doing a real implementation of these?

[xwu]

I didn’t invent these custom overloads, but my understanding is that they’re meant to provide a fast path when an element is clearly contained or not contained in the sequence.

As you know of course, in the general case, computing whether striding lands on any particular value between start and end is significantly more involved. I’m just committing a correctness fix here. Whether there are other scenarios that could benefit (performance-wise) from being quickly ruled in or out within this function as opposed to falling back can be left for future exploration.

[stephentyrone]

Yeah, I'm not asking you to solve the general case, I'm just surprised that it was implemented this way (and that there's no explanation or TODO to be found). I would appreciate if you added a note to that effect.

[xwu]

Yeah, I'm not asking you to solve the general case, I'm just surprised that it was implemented this way (and that there's no explanation or TODO to be found). I would appreciate if you added a note to that effect.

Sure thing—since the tests have mercifully completed successfully I'll follow up in a subsequent PR.

[xwu]

@stephentyrone Worth nominating for 5.5?