Better error checking

[fingolfin]

In a very naive attempt to try to use ferret to compute a normalizer), I entered this command (where G and H are permutation groups, and the goal was to compute N_G(H)).

gap> K:=Solve([ConInGroup(G), ConStabilize(Elements(H), OnSets)]); time;
Invalid attempt to read int
Error, Record Element: <rec> must be a record (not a boolean) in
  if record.generators = [ () ]  then
    retgrp := Group( () );
    StabChainMutable( retgrp );
else
    retgrp := Group( record.generators );
    SetStabChainMutable( retgrp, buildStabChain( record.generators, record.generators_map ) );
fi; at /Users/mhorn/Projekte/GAP/repos/pkg/ferret/lib/yapb.gi:526 called from
<function "Solve">( <arguments> )
 called from read-eval loop at *stdin*:188
you can replace <rec> via 'return <rec>;'

I believe the original source of the error is this line, resp. the lack of error checking before it:

  maxpoint := Maximum(List(constraints, x -> x.max));

The value of maxpoint ends up being a permutation.

So in retrospect, I understood that what I did was dumb. Still, I think ferret should catch this misuse, and report something sensible to the user.

Moreover, it seems that YAPB_SOLVE may return fail -- while ideally, one would catch all invalid input before invoking YAPB_SOLVE, that may not always be possible. In that case, fail could still be returned -- and hence I think it should be detected, and handled, instead of blinding treating its return value as a record.