Fix the Spline node to maintain stable segment IDs

[0HyperCube]

Closes #3318

[TrueDoctor]

I lack the mental model to judge if this is a good approach. Could this lead to cases where we merge lists of segments with the same ids? And would that create issues? If the change is conceptually sound, the code looks good, I just don't know how we would expect this to behave

[0HyperCube]

@TrueDoctor

The behaviour of producing small incrementing segment IDs is used by other nodes that generate geometry such as the circle and rectangle nodes.

In the flatten_path node, if there is a duplicate segment ID, it will deterministically modify one of the IDs through hashing (current_index, vector_index, node_id) and the previous segment ID:

Graphite/node-graph/gcore/src/vector/vector_nodes.rs

Lines 1018 to 1022 in 1b9a492

	let mut hasher = DefaultHasher::new();
	(current_index, vector_index, node_id).hash(&mut hasher);
	let collision_hash_seed = hasher.finish();
	output.element.concat(other, transform, collision_hash_seed);

Then generating the new IDs in the concat function:

Graphite/node-graph/gcore/src/vector/vector_types.rs

Lines 415 to 421 in 1b9a492

	let segment_map = additional
	.segment_domain
	.ids()
	.iter()
	.filter(|id| self.segment_domain.ids().contains(id))
	.map(|&old| (old, old.generate_from_hash(collision_hash_seed)))
	.collect::<HashMap<_, _>>();

This means that segment IDs must be:

• Unique within the current SegmentDomain.
• Deterministic between several runs of the same graph.

[TrueDoctor]

Alright, this looks good then, thanks for the detailed explanation @0HyperCube