BUG: Bravais determination eps fix
This mostly fixes a problem where Bravais lattice determination with coarse eps
might not produce the correct, more symmetric results, because the Niggli would use a smaller eps
. Instead it would recognize the less symmetric cells.
To totally fix it we must be able to guarantee that the Niggli/recognition eps
are somehow consistent. This may be the case now, but I don't quite know.