Account for mktemp failures

[ericonr]

Description

Discussed in #7477, catches when mktemp errors out and exits the script / function.

I can try to add some error case that exports TMPDIR to some place that can't be written, if you find it necessary.

TODOs:

• Changes to fish usage are reflected in user documentation/manpages.
• Tests have been added for regressions fixed
• User-visible changes noted in CHANGELOG.rst

[mqudsi]

Why don't we just add a tests function called mktemp? We need to do that anyway to get the tests to pass on older macOS versions that don't support the same syntax that Linux and newer macOS do.

[faho]

Why don't we just add a tests function called mktemp?

We would still have to add or return 1 everywhere, because the function won't return out of the calling function.

We need to do that anyway to get the tests to pass on older macOS versions that don't support the same syntax that Linux and newer macOS do.

As far as I'm aware, all these uses work on macOS and linux - the issue was stuff like too short templates or using templates with files instead of directories and such.

[ericonr]

I don't really have a horse in this race, I'm only extending the patch from my previous PR :)

So I'm happy to proceed with whatever's decided.

[faho]

Okay, I merged this because it's a good thing by itself. Making a compatibility function can be done later, and isn't immediately necessary as far as I can tell.

[mqudsi]

Agreed. Thanks, all!