Don't transfer nice monos when conjugating

[fingolfin]

We tried this but then decide against it, because we determined that often you conjugate precisely so that you get a nicer description of a group (see also Thomas' comment on PR #6183. But by accident the code implementing it got merged as part of PR #6183. Oops

Like PR #6206, this also fixes #6205 :-).

[lgoettgens]

(see also Thomas' comment on PR #6183

I understand that comment, that the nice mono should not be used when conjugating. On the other hand, #6183 conjugates the mono.

But I am fine with it either way

[ThomasBreuer]

The change is o.k., perhaps we should add a comment in the code why we do not transfer the known NiceMonomorphism value.

[ThomasBreuer]

Perhaps the formulation

this also fixes #6205 :-).

can be misunderstood.

The code from #6183 that stores a NiceMonomorphism value is mathematically correct, in this sense the currently proposed changes do not fix the bug behind #6205.

However, not storing the NiceMonomorphism has the effect that IsomorphismPermGroupForMatrixGroup cannot use it in a wrong way, hence the problem does not show up in the example from #6205.

[fingolfin]

Perhaps the formulation

this also fixes #6205 :-).

can be misunderstood.

Agreed... but in the most literal reading, I think it is appropriate: that issue is "semigroups tests started to fail", and this is fixed by this PR.

The code from #6183 that stores a NiceMonomorphism value is mathematically correct, in this sense the currently proposed changes do not fix the bug behind #6205.

Yes and no. I think there are at least two different issues behind #6205, and I made two PRs to address them.

This PR fixes the failures reported in #6205 because the code removed here caused a regression in the quality of the nice monomorphism being computed: instead of a perm rep on 8 points we suddenly produced one on 80 points. This regression was in two ways lucky:

1. it revealed another actual bug (good luck leading to PR Fix two bugs in IsomorphismPermGroupForMatrixGroup #6206) and
2. because of that, we detected the regression and were able to resolve it (good luck leading to this PR)

However, not storing the NiceMonomorphism has the effect that IsomorphismPermGroupForMatrixGroup cannot use it in a wrong way, hence the problem does not show up in the example from #6205.

Yeah but it shows up in other, harder to measure ways, in that some computations will perform worse, etc.