BUG: Prevent VOID_copyswapn ignoring strides

[eric-wieser]

Previously, this would cause a segfault in .sort.

Fixes #8210

[charris]

LGTM once the implicit declaration problem is fixed. I've often thought it would be good if we actually declared all functions at the top of the file before gettin around to the definitions.

[eric-wieser]

Fixed in a different way - having these functions call each other is a recipe for bugs if we end up modifying one and accidentally affecting the other. Better just to pull out some helper functions.

I think this first commit alone should be enough to backport, if you want to not do so to MAINT commits

[juliantaylor]

needs static

[eric-wieser]

@juliantaylor: Done

[juliantaylor]

can you squash the commits?
I don't like having known broken commits in the history for bisection.

[eric-wieser]

@juliantaylor: I'm not sure the first commit is broken? I think the tests just got cancelled

[eric-wieser]

It looks like I managed to confuse travis though, as the later tests are definitely running with a different set of source code to the early ones

[juliantaylor]

hm looks like it. Its fine then thanks

[eric-wieser]

Alarmingly, travis doesn't even have the same sha1 as the commit that cross is by