Added smoothstep built-in function

[Chaosus]

Personally I misses hermite interpolation. So many algorithms are depends on it. It should be in the core, cuz its common in video game programming. Unity also have it in the builtin Math class https://docs.unity3d.com/ScriptReference/Mathf.SmoothStep.html
In Godot, currently it can be only used in shaders(smoothstep) or in Tween's. This PR exposes it to core - I named it smooth_lerp for easy search it within other lerp-functions and(I suppose) its just another type of linear interpolation.

I will add C# version later, (if this will be accepted)

[bojidar-bg]

Lerp is a contraction for linear_interpolation. Maybe it should be called smoothstep/smooth_step?

[Chaosus]

Renamed to smooth_step

[Chaosus]

Added C# version too

[Calinou]

How does lerp() combined with ease() compare with this?

[neikeq]

C# change looks good to me but is wrongly formatted.

[Chaosus]

@Calinou explain more please ? I think its different functions, even if you somehow force the results to be equal its not as easy as single call

[bojidar-bg]

@Calinou Tested with KmPlot, the closest ease + lerp combination should be:

lerp(a, b, ease(t, -1.6521))

Here are some plots of both, along with their derivatives. Notice that ease's derivative has a discontinuity at x=0.5, unlike smoothstep's derivative.

f: smoothstep	g: ease	(f - g) * 100

[Chaosus]

Added is_equal_approx check to avoid division by zero

Its fixed !

[reduz]

nice addition, I wonder whether it should be "smooth_step" or "smoothstep" like in glsl..

[Chaosus]

Changed to smoothstep (as also suggested by Calinou in IRC)

[akien-mga]

Is it a problem if it's included further up before MATH_DECTIME, as for core/math?

[Chaosus]

by @vnen

I believe this breaks compatibility if put here: it should be put at the end of the list. You can see in the doc changes that all constants got new values, so if someone is using a function from below, suddenly the functions will be changed and it'll break the code.

[akien-mga]

Thanks!

[hpvb]

Cherry-picked for 3.1.1.

[kleonc]

@Chaosus, actually this implementation of smoothstep is similar to inverse_lerp and not to lerp so this description is misleading. Also another difference (besides interpolating faster at the beginning and slower at the end) is that smoothstep clamps result to range <0, 1>. So this description probably needs an update.
(no idea if that's proper place for this note)