Use skrifa::NormalizedCoord to represent normalized coordinates

[nicoburns]

The motivation for this PR is to avoid this conversion boilerplate (and otherwise unnecessary Vec allocation) when integrating parley (which uses swash's NormalizedCoord) with vello (which uses skrifa's NormalizedCoord).

Once parley is updated with this change it should be possible to pass run.normalized_coords directly to draw_glyphs().normalized_coords.

This PR assumes that the bit representation of normalized coords is the same in Swash and Skrifa, and that Skrifa just has a newtype around the raw bits.

[dfrg]

This looks reasonable and makes sense. The one downside is that updating the skrifa version is now potentially a breaking change but the interop benefit is probably worth it.

[nicoburns]

The one downside is that updating the skrifa version is now potentially a breaking change but the interop benefit is probably worth it.

Hmm... true. And I guess this is also currently true of Vello atm. Perhaps it would make sense to go the other way and use i16 as the interchange type?

This type is a re-export of F2Dot14 from font-types. So I think a skrifa bump would only be breaking if it bumps font-types. But still, we could otherwise make it entirely internal, which might be preferable?

[dfrg]

The other option is that vello’s DrawGlyphs could take an iterator instead of a slice. Doesn’t avoid the conversion but doesn’t require the allocation.

[dfrg]

I believe vello also currently re-exports skrifa so an update there is a breaking change regardless. Maybe we should rethink that.

[nicoburns]

The other option is that vello’s DrawGlyphs could take an iterator instead of a slice. Doesn’t avoid the conversion but doesn’t require the allocation.

That seems sensible anyway given that it only needs an iterator. But I was also thinking about Vello's release cycle. It would also be nice if a Skrifa bump in Vello didn't require a breaking release. Really the font-types crate ought to be serving this purpose of being the stable base (with infrequent breaking changes) that other crates can build upon. But I guess it's not quite there yet (but might be eventually?)

[dfrg]

The font-types crate has less churn than the others but it hasn’t really stabilized yet. Maybe in a few months.

In the meantime, we should consider changing vello to take an iterator, maybe of a simpler type like the current one in swash.

edit: and removing everything else that might make a skrifa update a breaking change.