Calling reduce_parity() in to_parity() is confusing, because then
running to_parity() on one SCC does not necessarily produce the same
output as running to_parity() on the entire automaton.  However it is
necessary for the implementation of parity_prefix.  As a compromise,
disable reduce_parity() when parity_prefix is disabled, this way we
can use that to demonstrate how the algorithm works.

* spot/twaalgos/toparity.hh, spot/twaalgos/toparity.cc: Do not
call reduce_parity() when parity_prefix is disabled.
* tests/python/toparity.py: Adjust.