string is pure ASCII: use _PyBytesWriter_WriteBytes(), don't check individual
character.

Cleanup unicode_encode_ucs1():

* Rename repunicode to rep
* Clear rep object on error
* Factorize code between bytes and unicode path