Allow nicer access to underlying Julia object for wrapped Julia functions

[fingolfin]

We already have UnwrapJuliaFunc. But it would feel more natural to me if we (also?) had a JuliaPointer method applicable to them.

Bonus points for also putting them into filter IsJuliaWrapper but that's more tricky and likely requires modifying the TypeObj callback table for T_FUNCTION

[fingolfin]

The GAP kernel does this:

static Obj TYPE_FUNCTION;
static Obj TYPE_OPERATION;
static Obj TYPE_FUNCTION_WITH_NAME;
static Obj TYPE_OPERATION_WITH_NAME;

static Obj TypeFunction(Obj func)
{
    if (NAME_FUNC(func) == 0)
        return (IS_OPERATION(func) ? TYPE_OPERATION : TYPE_FUNCTION);
    else
        return (IS_OPERATION(func) ? TYPE_OPERATION_WITH_NAME : TYPE_FUNCTION_WITH_NAME);
}

[...]

    TypeObjFuncs[ T_FUNCTION ] = TypeFunction;

So we could do something like this, roughly:

static Obj (* OriginalTypeFunction)(Obj obj);

static Obj TYPE_JULIA_FUNCTION;

static Obj TypeFunctionJulia(Obj func)
{
    if (IS_JULIA_FUNC(func))
        return TYPE_JULIA_FUNCTION;
    else
        return OriginalTypeFunction(func);
}

....

// inside a suitable init function
    OriginalTypeFunction = TypeObjFuncs[ T_FUNCTION ];
    TypeObjFuncs[ T_FUNCTION ] = TypeFunctionJulia;

Alternatively, instead of the indirect jump (which is slow) we could of course also just duplicate the original logic.