Refactor GraphEdit connections

[Jummit]

There was code to draw bezier curves in graph_edit.cpp, which I removed to use Curve2D instead.
This PR makes it possible to retrieve the points of a connection line between two given points, and to customize how connections are drawn:

Examples of custom connections

func _get_connection_line(from: Vector2, to: Vector2):
	var distance = to - from
	return [from, from + Vector2(distance.x / 2, 0), to - Vector2(distance.x / 2, 0), to]

func _get_connection_line(from: Vector2, to: Vector2):
	var distance = to - from
	return [from, to]

When _get_connection_line isn't overriden:

By adding a way to access the connection line, it makes it possible to implement interaction, like putting nodes in between or adding reroute nodes:

Gist:
https://gist.github.com/Jummit/545d193b9f5119a428862abfefe4be53

Issues to fix:

• not tested on the latest build, as my computer can't run Vulkan
• Colors don't work
  Example

  
  

• I'm not sure about the get_connection_line name, as it's very close to get_connection_list.
• inconsistent with the old line when the connection goes backwards (or is this ok?)
  Example

  
  
  Slide Comparison: https://imgsli.com/MTkyOTc

[swarnimarun]

I wanted to do it soo badly.
You have my thanks.
(Now if someone adds better styling support for graph/graphnodes and then zooming(I mean finish the PR), I will be soo happy that I will scream... :p)

Once it's done I will see if I can find time to add these features to VisualScript and VisualShader Editor.

Will review it once I have some free time.

I have like 2 questions,

1. How does it handle overlapping?
2. What do you mean by wrong result when backwards.?
   How is it affected by overloading get_connection_line?

Also you can do, "get_connection_spline" instead more technically accurate I think.

[Jummit]

I wanted to do it soo badly.
You have my thanks.

No problem.

1. How does it handle overlapping?

Like the old system, there should not be any changes. It still uses draw_polyline_colors, just with different points.

2. What do you mean by wrong result when backwards.?
   How is it affected by overloading get_connection_line?

It is inconsistent with the old look:

https://imgsli.com/MTkyOTc

Tell me if that's fine, or if it should be made to look more like the old line.

Also you can do, "get_connection_spline" instead more technically accurate I think.

It's only a spline if you use a Curve2D (the default), but you can add any line, as shown in the examples.

[akien-mga]

Needs a rebase on current master to fix CI.

[swarnimarun]

Also you can do, "get_connection_spline" instead more technically accurate I think.

It's only a spline if you use a Curve2D (the default), but you can add any line, as shown in the examples.

I mean splines can be linear as well, spline represents a set of line segments between the two adjacent data points

Aside: If I am wrong pls correct me. :3

[Jummit]

Needs a rebase on current master to fix CI.

Done.

[fire]

I and @swarnimarun will review on a spare moment.

[fire]

I'm looking at the code and the resulting lines are too thin.

It seems to be related to

p_where->draw_polyline_colors(points, colors, 2 * EDSCALE);

However, I don't remember where the default for this is.

[fire]

RenderingServerCanvas::canvas_item_add_polyline has not implemented width. I'll have to circle back to that.

[aaronfranke]

@Jummit Is this still desired? If so, it needs to be rebased and tested on the latest master branch.

[Jummit]

I'm hesitant to work on it because I can't test the master branch, and there are some graphical inconsistencies I would need to look into. So closing this, but it's pretty salvageable if anyone would like to pick it up. If I ever get a PC that supports Vulkan I'll rebase and test it.