Fixes for `#pragma warning`

[Daniel-Cortez]

What this PR does / why we need it:

• Fixes #pragma warning disable not being reset at the end of each compilation pass, resulting in its effects being carried onto the subsequent compilation passes (see #610).
• Makes the compiler issue an error when #pragma warning push is used more times than #pragma warning pop and vice versa (see #610).
• Changes the return value type of pc_pushwarnings() and pc_popwarnings() from int to void. pc_pushwarnings() never returns FALSE; in case of memory allocation failure it calls error(103), which aborts execution (as 103 is a fatal error), making the line return FALSE; unreachable. Also, the values returned by both pc_pushwarnings() and pc_popwarnings() are never used.
• Covers #pragma warning push/pop/enable/disable with tests (previously there weren't any).

Which issue(s) this PR fixes:

Fixes #610

What kind of pull this is:

• [x] A Bug Fix
• [ ] A New Feature
• [ ] Some repository meta (documentation, etc)
• [ ] Other

Additional Documentation:

[Daniel-Cortez]

#if defined ENABLE_ALL_WARNINGS
#pragma warning pop
#endif

It seems that you're completely missing the point on how #pragma warning push/pop works.

Firstly, #pragma warning pop doesn't enable all warnings, it reverts them to the state previously saved with #pragma warning push.

Secondly, if you thought #pragma warning pop would simply enable all warnings if there was no matching #pragma warning push, then again, no, it doesn't do anything in such cases. For example, try compiling this code with the v3.10.10 release:

#pragma warning disable 203
#pragma warning pop
new var;
main(){}

In this example the compiler wouldn't warn about var being unused, because warning 203 is disabled and #pragma warning pop doesn't actually do anything when the warnings stack is empty (no prior #pragma warning push).

Thirdly, usually an unmatched pop is a sign that the user forgot to use #pragma warning push somewhere, which is why I made the compiler issue an error in such cases, to inform the user about their mistake instead of silently ignoring the problem.

[Y-Less]

TBH I was unsure about some of these things, but pop means "restore the last item". If there's no item to restore, what should happen?

[Y-Less]

Nothing wrong with complex sets of conditional pushes/pops, as long as they match up in the end.

[Y-Less]

If you don't want to pop, just don't push. Disabling a warning doesn't require a push first.

[Y-Less]

But that code is probably a mistake. Maybe it should be a warning instead, but still. The pop will only restore the most recent push, but there are three pushes and three warning disables there. So in that code after the pop only warning 215 will be restored. You'd have to do:

#pragma warning pop
#pragma warning pop
#pragma warning pop

To restore all three.

[Daniel-Cortez]

Maybe it should be a warning instead, but still.

Well, as I already mentioned in #610, the main reason I made them as errors was because I wanted unmatched #pragma warning push/pop to follow the same rules that already apply to unmatched #ifs and #endifs. I wouldn't mind changing them into warnings though, but that would require adding at least 1 new warning for this, and there's already a PR that implements a new warning (614), which probably needs to be merged in first in order to avoid possible merge conflicts.