Moving variable font logic into `read-fonts`

[LaurenzV]

I was trying to look into whether I can use read-fonts to build an instancer for TTF fonts, for usage in PDFs. Building a full-blown instancer is obviously a huge task, but luckily, for PDF, I only need to instance the hmtx and glyf tables, which I'm hoping will be much more doable.

I still need to do some proper reading to assess what I need to do to achieve this, however, at a first glance, when looking into this I noticed that most of the VF handling for glyph outlines seems to be inside skrifa in deltas.rs. Would it make sense to move some of this into read-fonts, so that other crates can re-use the code, or is there a particular reason why it's not in read-fonts? Another option would be to expose it in skrifa, but it seems to me like this would go against its philosophy, since skrifa is supposed to be very high-level and it also would be unfortunate to have to depend on skrifa just for that.

[dfrg]

I don’t see any problem with moving this to read-fonts, perhaps as methods on the gvar table. The API requires passing in a lot of buffers (to avoid internal allocation) so that might need to be cleaned up and better documented for public use.

[rsheeter]

Building a full-blown instancer is obviously a huge task

FYI, we do plan to tackle this, likely after klippa (memory safe hb-subset) wraps up.