Cleanup dithering conditional and explanation

[OrangeChannel]

Shouldn't change anything because calling resize with float output and a dither_type disregarded the dither_type value anyways. Reverts some of the changed behavior in #39 and avoids the misleading should_dither being true for float sample type.

[stuxcrystal]

We are at our third PR with discussion about this topic iirc.

At this point I think we should split the one-liner into multiple lines and put them into a (possibly internal) function.

[OrangeChannel]

This can get squashed into something like fix dithering logic (#41) and tagged as 0.3.1. I'll then rebase #37 ontop of this since it shares some of the test code and depth needs the decorator too.

Or merge #37 first (tagged as 0.4.0) and then I'll rebase this and get this tagged as 0.4.1 (which should be the next pypi/package release).

[kageru]

The tests are fine, but _should_dither is way too enterprise. Just make it a wrapper around the boolean condition that we had before, optionally with some nicer names like you already did here, but don’t give it all those pseudo-optional parameters where only some combinations of them are actually valid.