Sync math defs and functions with upstream Godot

[Togira123]

This PR syncs math functions and definitions with upstream Godot.
Notable changes include adding constexpr to many functions and changing macro definitions to be constexpr constants in the Math namespace. Some math function bodies were changed to stay in sync with upstream godot and to allow the constexpr specifier.

[Ivorforce]

This looks good. Am I right in assuming you've worked manually (adding constexpr to the existing bodies)?

The way we often sync looks more like "copy-paste the upstream version, then review the diff to godot-cpp-ify it again".
For example, it seems like you're skipping SQRT3 and upstream updates to smoothstep and abs.

So I think this is mergable/a clear improvement, but I wouldn't consider it to be a full sync yet.

[Togira123]

What I did was manually compare the functions to the upstream version and made changes where needed. Copy-pasting seemed a bit more unreliable here since I noticed that we have functions that upstream does not have and many functions that upstream has weren't synced in the past either, additionally functions are in a different order making reading diffs properly harder. I can still take another look and fix the remaining changes though yes :)

[Ivorforce]

Sounds good!

[Togira123]

Oh and another thing. I saw that our Math::sign implementation uses SIGN from defs.hpp, while min, max and clamp do not. Should I change these three to also use the respective implementations from defs.hpp for consistency?

[Ivorforce]

Good question. It's interesting that godot-cpp has versions of them in the first place even though we usually try to avoid adding more features that are not in godot upstream.

I'd say yes, keep them to avoid compat breaks (and because godot might move them to Math eventually too), but use the canonical defs.hpp implementations.

[Togira123]

Noticed that there is a mix of using static_cast and C-style casts in math.hpp (in places where static_cast would also be applicable). I can replace C-style casts with static_cast where possible if you want.

[Ivorforce]

Whatever upstream Godot uses is fine. static_cast is indeed preferable, but the more we deviate from upstream, the harder it gets to synchronize in the future.

[Togira123]

There are functions that weren't synced before. My assumption is that that's because they aren't exposed in gdscript to keep the two APIs similar. Should I continue only adding functions that are available in the latest godot version's gdscript?

[Ivorforce]

I'm not aware of the history, but I'd just sync everything. The goal of godot-cpp is to match Godot's own c++ as closely as possible, not necessarily the GDScript API.

[Togira123]

Alright, a lot has changed😅 comparing math_defs.hpp with upstream now it is the same except for the enums, which I saw are generated automatically, hence I didn't include them. math.hpp is also almost identical to its godot counterpart, except for a few function declarations that weren't synced before either.

[Togira123]

removed the int_alt_t and uint_alt_t since std::conditional_t was causing compilation errors on linux and windows. Not sure whether we need these two anyway

[Ivorforce]

Whew, this turned quite large now, but going by the diffs it looks all good!

For posterity: Math_PI etc. macros, absf/absd, db2linear/linear2db are all deleted, and CMP_EPSILON etc. move into the godot namespace. All without no compatibility shim.
I don't think any of these are widely used, they're easy to fix, and providing no compat is defensible for 10.0. So I agree with the decision.

Thank you for taking care of this!!