Simplify Isometry multiplication benchmarks (#2628)

With the benchmark library, there is no need to specify an iteration count.
Interestingly, 4x4 matrix multiplication is faster than affine*matrix

Co-authored-by: Mario Prats <mario.prats@picknik.ai>