Fix mapped int negative or overflow bugs, improve color C API

[ankith26]

The goal of this PR is to fix #2336 and fix #1408 (the original)

The core issues are that a color is actually Uint32, but long on windows is actually Sint32. This leads to

• Functions returning a mapped int can return a negative value (values above and including 0x80000000 are casted incorrectly when they are Sint32)
• Functions expecting a mapped int can sometimes handle this negative value
• Functions expecting a mapped int can sometimes fail to handle a proper mapped int above 0x80000000

The route I'm taking in this PR that involves least user-facing changes and breakages is

• No changes to functions returning mapped ints (maybe this could/should be changed, but I'd like to defer that to a future PR)
• All functions handling mapped int passes through the same internal C API, and this should handle both negative and positive-but-bigger-than-0x80000000 values, and also cap further overflows and error correctly

In the process, I figured color handling API was in a bit of a mess. We don't handle str and int colors in a lot of places. But in this PR I did not want to change anything that can have a user-facing impact (apart from the bug). This bug is related to issues in the python 2 to 3 port.

I introduced unified API for color handling (in a way that doesn't change anything userfacing)

Here is a list of old API, with new API equivalents

• pg_RGBAFromColorObj(...) (now removed) -> pg_RGBAFromObjEx(..., PG_COLOR_HANDLE_SIMPLE)
• pg_RGBAFromObj(...) -> pg_RGBAFromObjEx(..., PG_COLOR_HANDLE_SIMPLE)
• pg_RGBAFromFuzzyColorObj(...) (now removed) -> pg_RGBAFromObjEx(..., PG_COLOR_HANDLE_ALL)

A new addition pg_MappedColorFromObj replaces copypasted block of code that handled int incorrectly (this is what fixes this bug)

The final goal is to get rid of these temporary flags entirely, as every function should handle all ColorValues. But that involves writing more docs and tests, so it's coming in a future PR.

[zoldalma999]

Overall, I like this pr, makes handling color-like objects easier. Left a few questions.

Also, when we planning to change PG_COLOR_HANDLE_SIMPLE calls to handle everything?

[MyreMylar]

For whatever reason CircleCI here is trying to build with python 3.6 - which we don't support - so it is failing.

[MyreMylar]

LGTM 👍

The new API makes sense, and the color tests run locally for me. I also couldn't find any lingering usages of the old ways.

[dr0id]

removing code duplication is always good and makes handling colors more consistent.

LGTM