MatrixF changes

With Marcelo's fix for the notorius "pink screen" bug for users of machines that do not have SSE support, some changes have occured to MatrixF that you guys should be aware of.

MatrixF::mul will assert if you attempt to call it with parameter that includes the "source" matrix. Ie. this is invalid:

MatrixF foo, bar;
foo.mul( bar, foo );

The reason is that it will attempt to multiply bar by foo while it's overwriting foo. Rather than handle this case internally in the function, we assert (to keep MatrixF::mul speedy).

OK, fine, but now you might want to know how the @#$* you are supposed to perform a pre-matrix multiply without explictly allocating a new matrix. The answer is we now offer operator overloads for * and *= with matrices. So now you can do:

foo = bar * foo;

or

foo *= bar;

or

foo = bar * foo * bar;

But note that

A.mul( B, C );

is faster than:

A = B * C;

On the other hand

A.mul( B );

is the same speed as:

A *= B;