matrices vs. matrix objects: remove two implications

[ThomasBreuer]

This is an experiment mentioned in #6221:

• Remove the implication from IsGF2MatrixRep to IsRowListMatrix, such that IsGF2MatrixRep no longer implies IsMatrixObj.
• Analogously, let IsGF2VectorRep not imply IsVectorObj.

The first aim is to see how many package tests fail due to this change.

[fingolfin]

I've triggered a test run of the package distro tests against your PR.

[ThomasBreuer]

Thanks @fingolfin .

[fingolfin]

This revealed failures in fining, forms, and semigroups. I did not yet look closer.

[ThomasBreuer]

Concerning the failures of package tests:

• Semigroups:
  In gap/semigroups/semigrp.gi, the check IsMatrixObj(Representative(gens)) no longer returns true for the representative in IsGF2MatrixRep or Is8BitMatrixRep. As a consequence, the filter IsMatrixOverFiniteFieldSemigroup is not set in the semigroup in question, and then no BaseDomain method is applicable.
  A solution would be to check for IsMatrixOrMatrixObj(Representative(gens)) instead.

• Fining:
  This points to a problem in the cvec package:
  Its NewMatrix functions call (the two argument version of) CVec in different ways, depending on the representation of the entries of the given list.
  For IsPlistRep, nothing is to do, and IsVectorObj requires an Unpack; this Unpack is needed for "proper" IsVectorObjs but apparently also for IsGF2VectorRep and Is8BitVectorRep.
  A solution would be to add the handling of IsGF2VectorRep and Is8BitVectorRep analogous to IsVectorObj.

• Forms:
  This problem disappears as soon as add ConformalSymplecticGroup and some generalizations gap-packages/forms#80 gets merged.

[ThomasBreuer]

This proposal is not acceptable for the Semigroups package, see the comment in #6221.