- expose ReferenceError
- remove duplicate fg/bg chans

- we can now attach "broken" flags to apps
- we can now provide patches for "broken" apps without
  writing to original repo while being transparent to user
- users can now look for updates for installed apps
- app writers can define color schemes for their app store
  pages in flow3r.toml
- description text is displayed on app store page
- more colors!

see previous commit.

It is more descriptive like this probably, and this way we can
add the angles as radiants too to avoid inefficient cmath.phase
calls or user side caching.

For all other methods the beginner's rule of thumb is "when in doubt,
call `super().methodname()` when overriding, but for draw it gives a
cryptic @abstractmethod error:

File "/flash/sys/st3m/ui/view.py", line 163, in draw
TypeError: function takes 0 positional arguments but 2 were given

This is not very beginner friendly, so we're providing a NOP default.

This is due to a subtle change in behavior of our duck
_scroll_controller, but it's a good one so we'll keep it.

There was a performance regression since then, probably located in these upstream commits:

* 720a5357 "rasterizer: always do at least one subidvision of beziers" - 15% FPS drop on sensors app
* 5b4d3d44 "ctx: the glyph API now takes glyph_ids and not unicode codepoints" - further 25% FPS drop on sensors app

Not terribly sure about the last commit because it doesn't work by itself (most glyphs don't render, this was fixed later in b5a9a81d), but it's most likely the group started by that one.

These commits happened to be 2-3 days after our last known good commit, and performance of the sensors app remained stable for months afterwards.

Because we're preparing for a release and pippin doesn't seem to have the time to address this right now, and I'd have no idea where to even begin, I'm doing the next best thing I know and reverting to the known good version.

They have lower bulk capacity, presumably due to speaker holes,
and work just fine.