Replace many uses of G.X by GapObj(X)

[fingolfin]

... so that ultimately we can refactor the underpinnings of GAPGroup etc. -- initially by dropping those slightly evil getproperty methods that produce G.X on demand.

In the future this enables other refactoring. E.g. we could contemplate adding a mechanism where one could replace GAPWrap.Foo(GapObj(x)) by GAPWrap.Foo(x) (namely: whenever the wrapper specifies that an argument must be a GapObj, as in e.g. GAP.@wrap ClassNames(x::GapObj)::GapObj -- then the generated wrapper may as well assume it has to convert anything else to GapObj.

Anyway: this PR does not strive to eliminate all accesses to .X -- first off, some of them are of course legit. Secondly, this is already very big and I'd rather resolve a large chunk of the issue now, before delaying this further and having to deal with merge conflicts...

[ThomasBreuer]

Good.

Tests fail because GapObj is not known in the tests; the (few) tests that use GapObj already now call it as GAP.GapObj. Shall we better export GapObj, since now its use as a constructor in tests seems to be natural?

(It is interesting that GAPGroupConjClass has a field X whose value is the Oscar object given by acting_group, and that the underlying GAP object is stored in CC.)

[fingolfin]

Tests fail because GapObj is not known in the tests; the (few) tests that use GapObj already now call it as GAP.GapObj. Shall we better export GapObj, since now its use as a constructor in tests seems to be natural?

Seems reasonably, so I've now exported GapObj.

(It is interesting that GAPGroupConjClass has a field X whose value is the Oscar object given by acting_group, and that the underlying GAP object is stored in CC.)

"Interesting" is putting it nicely -- I found it very confusing and surprising, and actually additional motivation from moving away from access to FOO.X to use of accessor functions with more self-explanatory names :-)