Implement extra cursors on macOS

[mantognini]

Superseds #1413 & #1401.

@JonnyPtn when carefully reviewing your PR I noticed a few issues in my original code so I've fixed it here. Feedback welcome.

I've removed Wait and SizeAll cursors as they don't look nice (Wait is not spinning and produces a broken rendering, SizeAll is a simple white cursor as you mentioned.) I think it would be more a disappointment/confusion or a source of bug reports to keep those.

It should also be more robust w.r.t. memory handling and reduce unwanted selector warnings.

[eliasdaler]

Look at how SDL does it. Maybe we can do the same too?

https://github.com/spurious/SDL-mirror/blob/a4e95f81625b16d7454a7edb4010e7d37f72cebf/src/video/cocoa/SDL_cocoamouse.m#L136

I'm not sure about SDL_SYSTEM_CURSOR_SIZENWSE and SDL_SYSTEM_CURSOR_SIZENESW choice (as other resize* cursors work well for that!), but it's better to have closed hand as SizeAll then nothing.

[mantognini]

but it's better to have closed hand as SizeAll then nothing.

I'd rather let the users decide on this themselves than impose an arbitrary and imperfect solution on them. ;-)

[JonnyPtn]

Isn’t it an arbitrary and imperfect solution whether we used closed hand or leave it as normal arrow cursor?

I’d also say closed hand is better than unchanged

[mantognini]

The current choice is between the closed hand (or any non optimal cursor) and returning false to explicitly let the user choose another cursor.

[JonnyPtn]

Ah my mistake, returning false is more logical

[eliasdaler]

Agreed.

[eXpl0it3r]

Since this doesn't change any public-facing API or files, I've moved this from "feature" to "bugfix" and will include it in SFML 2.5.1.

@mantognini Let me know if I should rebase it or whether you want to do it. 🙂

[eXpl0it3r]

Merged in aeca3dc