postponed to 1.1

As discussed on a call, the cache is architecture-dependent so it has to be run on the target, or in a qemu.

I'm not convinced that the cache generation is sufficiently slow to make the overhead of qemu worth it.  I suspect consolidating the multiple cache generation operations into a single operation would be a better win.

(that's entirely without data, just a hunch)

So as the fontconfig cache is architecture dependent, so either
  - build it on the target
  - build it on the host in qemu

I'm not convinced that the overhead of qemu is worth the effort.  We used to build the locales this way because low-memory devices didn't have enough memory to the work.

I'll close this as WONTFIX, and open another bug about moving the font cache generation to a trigger.

Setting this to verified due to Ross latest comment