Add NaturalGModule for obtaining the natural module of a matrix group

[fingolfin]

Also fix usage error message for GModuleByMats.

The main gist is: instead of this

g:= SP(4, 2);;
m:= GModuleByMats( GeneratorsOfGroup( g ), GF(2) );;

you can now write this

g:= Sp(4, 2);;
m:= GModuleByGroup( g );;

which just is quite convenient when dealing with matrix groups over finite fields.

A field can be specified as an optional second argument in case the matrix groups is defined over one field, but we want the module over another. But perhaps I should just drop it?

This is a draft, as there is no documentation yet for the new function (just a placeholder), and also a TODO comment here or there in the code.

Before I fine tune it I wanted to check if this seems sensible. Also I wonder whether the function's name should be changed to e.g. just GModule, or perhaps NaturalGModule, or something else.

[ThomasBreuer]

Good.

I think NaturalGModule would be a natural name; this term occurs for example in the description of PermutationGModule.

If the function shall eventually cover also the case of permutation groups then the argument that specifies the field cannot be avoided; then perhaps PermutationGModule need not be documented anymore.

(Perhaps for another pull request: The check of arguments of GModuleByMats can be improved; a given dimension is ignored if the list of matrices is nonempty, and it is not checked that the matrices really fit together; for example, GAP accepts the input GModuleByMats([[[Z(2)]], [[1]]], GF(2)).)

[ThomasBreuer]

One of the tests is failing, a cross-reference should be adjusted.
(This is actually not an error in the sense of GAPDoc, but somehow more rules are imposed on the GAP manuals; unfortunately some of these rules are wrong.)

[ThomasBreuer]

Now the tests pass.