This PR has a couple of minor changes from #48:

1. Better error text for out-of-bounds coordinates
2. DecodeCharError renamed to EncodeToCharError since that's what we're trying to do when the error occurs
3. A new CoordEncodingError which is the "parent" of EncodeToCharError:

The reason for 3. is that we lack index or Coord information inside encode(). When an EncodeToCharError occurs there, we early-return back up to its calling encode_coordinates() parent function, where we do have index and Coord info, so we map the bare error onto the richer "parent".

This change has no perf implications in my testing.