Add properties to CharacterBody for more move_and_slide options

[fabriceci]

Implementation of the proposal 2982 that aim to fix and add more options to move and slide in 2D, see the proposal for details.

I made the following project https://github.com/fabriceci/move_and_slide_4.0 to test this PR easily.

[pouleyKetchoupp]

Great work!

I haven't had time to do extensive testing yet, but I've left some comments on the code and documentation. There might be a bit more later when I get deeper into the math :)

[pouleyKetchoupp]

I've made another batch of review comments about details of the algorithm. You changes generally look great, my comments are mostly about making things a bit more readable and adding comments to help future contributors.

I still haven't done the regression tests I wanted to, I'll probably do it next week after porting some example projects to 4.0.

[pouleyKetchoupp]

Here are the tests I've made so far:

• Test on ground and platforms
  (from https://github.com/fabriceci/move_and_slide_4.0)

No regression spotted, everything seems to work well.

• Test on curved floor
  snapping_testing-4.0.zip
  (converted MRP from 2D KinematicBody slides down slope with move_and_slide_with_snap #31097)

No regression.
Constant speed works well.

Possible issue: When setting motion on floor only (default move_max_angle) the character is able to climb a little bit on the curve at the 45 degree limit, and slide back a little bit, instead of just being stopped.
It's not a regression since it only affects the new settings, so not a requirement for this PR, but eventually it would be nice for curves to be handled with no glitch.

• Test with basic sliding
  using_kinematic2d-4.0.zip

Constant speed works well.

Possible regression: When setting move_max_angle to 180 degrees, the character is not able to climb the steeper slope anymore when moving right. It jitters on the edge instead. I would expect that it would work the same as before this PR in this case (the character was able to move on walls when their angle is low enough, and slide back when standing).

Before	With this PR

Edit: After discussion with @fabriceci and more tests, it makes sense that it's a case not supported by default. We can actually get rid of move_max_angle which is a confusing parameter (it allows walking on walls) and instead it's possible to change up_direction by script to achieve similar results. So the general idea can be that whatever is not floor is considered not walkable at all, it's much easier to understand and document properly.

Just for the record, here's a snippet of gdscript that can achieve something similar (before move_and_slide()):

if ((abs(linear_velocity.x) > 0) && is_on_wall() && get_slide_count() > 0):
	var last_collision = get_slide_collision(get_slide_count() - 1)
	var collision_normal = last_collision.get_normal()
	var collision_angle = abs(rad2deg(acos(collision_normal.dot(Vector2.UP))))
	if (collision_angle < 90.0):
		up_direction = collision_normal
		stop_on_slope = false
else:
	up_direction = Vector2.UP
	stop_on_slope = true

• Test with low-res platformer game
  simple--heart-platformer-master-4.0.zip
  (partially converted MRP from 2D platform physics collision jitter #45259, I had to redo a tilemap and the pixelated low-res settings don't work well in 4.0)

No regression.

Next I'm going to convert physics tests from https://github.com/godotengine/godot-demo-projects/tree/master/2d/physics_tests to 4.0 in order to do the last remaining regression tests.

[fabriceci]

Possible issue: When setting motion on floor only (default move_max_angle) the character is able to climb a little bit on the curve at the 45 degree limit, and slide back a little bit, instead of just being stopped.
It's not a regression since it only affects the new settings, so not a requirement for this PR, but eventually it would be nice for curves to be handled with no glitch.

I struggled a lot on this. Note: there was a small problem in the test project you used, the gravity was not applied on the ground.

[pouleyKetchoupp]

Just a few last things to adjust, otherwise it all works perfectly well!

Edit: Also tested with physics tests converted to 4.0, and no regression spotted on character controller tests.
(godotengine/godot-demo-projects#638)

[pouleyKetchoupp]

Amazing work, thanks!

[akien-mga]

Thanks!