Deprecate/remove MTI feature compiler?

[simoncozens]

Google's Noto project was almost certainly the only user of the MTI feature compiler (compiling Monotype FontDame text-format GSUB/GPOS dumps to binary). With all of the MTI format files now converted to FEA, this is no longer a requirement. While there's no harm having the old code around, and possibly some benefit in terms of separation-of-concerns in having two feature compiler subclasses, less code is cleaner and more maintainable than more code. We should deprecate and remove it.

[behdad]

some benefit in terms of separation-of-concerns in having two feature compiler subclasses

That's the only reason I have for liking it: to keep otlLib honest. But I'm fine removing it.

[simoncozens]

otlLib/feaLib/mtiLib is a fontTools distinction; mtiLib can stay in fontTools for that purpose. But we also have a FeatureCompiler and an MtiFeatureCompiler here in ufo2ft (and to a lesser extent in fontmake), and I think they don't gain us very much.

[anthrotype]

I'm ok to deprecate and eventually remove the feature from ufo2ft and fontmake.
The way it was implemented was always a bit awkward in that required that first the master TTFs be buillt with the first invocation of fontmake, then again a new fontmake call to interpolate the instances and the accompanying interpolated GPOS tables (--interpolate-binary-layout option). I doubt anybody else besides the old shell scripts in the noto-source repository ever used this.