[spv-out] Fix duplicate scalar OpType

[Napokue]

When the following WGSL input was used:

[[stage(vertex)]]
fn main() -> void {
  var vector: vec2<f32> = vec2<f32>(1.0, 1.0);
  return;
}

You would get a validation error when validating the SPIR-V output. The reason was because there would be a duplicate OpTypeFloat is because the vector implicitly uses OpTypeFloat here.

If you would first explicity add a float variable, it would work:

[[stage(vertex)]]
fn main() -> void {
  var float: f32 = 1.0;
  var vector: vec2<f32> = vec2<f32>(1.0, 1.0);
  return;
}

The reason this works is as there is an explicit f32 declaration and that is getting parsed correctly.

Technical

This is due to how LookupTypes work and I am not entirely sure how it is caused, but it something has to do that the implicit f32 don't have a handle and aren't recognized when looking up.

Conclusion

For now, this "quick" fix should push the code forward, so we can focus on getting in more support for more shaders. The Lookup system works fine for now, but it can use some refactor a little later down the road.

This PR is one step closer to successfully parse the quad.wgsl

[kvark]

Thank you for identifying and fixing this! I believe the fix is in the right direction.
The whole lookup_type set of guarantees is a bit messy still, since it's modified in multiple places, but refactoring this has to wait for a better time.

[kvark]

ok, postponing the naming changes