Implement proper fill rendering for vector meshes

[liunicholas6]

Closes #3378
Didn't want to have to deal with working off of @iamthenoname repo so I made a new set of commits.
@0HyperCube Thanks for all the feedback!

[Keavon]

!build

[liunicholas6]

Not sure what the process is for adding dependencies -- I saw fixedbitset is a dependency of petgraph so I thought it would probably be OK to bring in; I can easily swap it out for a Vec:: or similar

[Keavon]

@TrueDoctor can you give your thoughts on this please?

[github-actions]

📦 Build Complete for 8bdc906
https://b61a599b.graphite.pages.dev

[liunicholas6]

Update -- looks like some fills still aren't working; probably going to take a lot of debug logging to figure out. Will take another look tomorrow

[0HyperCube]

If the handle is placed on top of the anchor then this will be zero.

For reference, here is my impl of the tangent at start which attempts to avoid reporting zero length tangents:

Graphite/node-graph/libraries/vector-types/src/vector/algorithms/bezpath_algorithms.rs

Lines 704 to 724 in 06749ec

	/// Compute the tangent at t=0 for the path segment
	pub fn tangent_at_start(segment: kurbo::PathSeg) -> kurbo::Vec2 {
	let tangent = match segment {
	kurbo::PathSeg::Line(line) => (line.p1 - line.p0).normalize(),
	kurbo::PathSeg::Quad(quad_bez) => {
	let first = (quad_bez.p1 - quad_bez.p0).normalize();
	if first.is_finite() { first } else { (quad_bez.p2 - quad_bez.p0).normalize() }
	}
	kurbo::PathSeg::Cubic(cubic_bez) => {
	let first = (cubic_bez.p1 - cubic_bez.p0).normalize();
	if first.is_finite() {
	first
	} else {
	let second = (cubic_bez.p2 - cubic_bez.p0).normalize();
	if second.is_finite() { second } else { (cubic_bez.p3 - cubic_bez.p0).normalize() }
	}
	}
	};
	debug_assert!(tangent.is_finite(), "cannot round corner with NaN tangent {segment:?}");
	tangent
	}

[liunicholas6]

Yep -- I special-cased tangent_at_start and tangent_at_end for each type of segment but then didn't actually call into them on the match case.

I didn't think of the "find the first non-overlapping control point" strategy though; that's a good call that I addressed on revision.

[Keavon]

!build

[github-actions]

📦 Build Complete for 1c733bb
https://45259c6a.graphite.pages.dev