In BLAS' funtions, dgemm allows to make C <- a*A*B + b*C, but MuJoCo doesn't allow to scale the A*B multiplication in just one shot. I guess it could be a good thing to include a scale factor in the mju_mulMatMat function and its brothers.
MuJoCo's utility functions are not meant to replace BLAS. The design principle is that any functionality which is used internally is exposed to the user, but no additional functionality is implemented. The alternative would be to implement all of BLAS, which makes no sense. So just link a matrix library of your choice, and keep in mind that MuJoCo's matrices are in row-major format.
Ok. I undestand. Although the scaling factor could makes things easy (and faster) sometimes, the implementation of all BLAS indeed doesn't make sense. And thanks for the row-major information. That's important.