cbprintf_packaged: code cleanups

- Abstract buffer offset computation for better code clarity.

- Rework the logic around rw/ro strings to simplify the logic and
  to guard against overflows even when only computing the needed buffer
  size.

- Use modulus to simplify alignment tests (generated assembly is
  the same).

- Avoid CBPRINTF_ prefixes for local macro names

- Better pointer types to reduce cast usage.

- Add more comments.

Signed-off-by: Nicolas Pitre <npitre@baylibre.com>