[f2py] control number of OpenMP threads at runtime

Hi

I use OpenMP to speed up subroutines in my extension modules. I don't
set the number of threads in the code (e.g. with !$omp parallel
num_threads(N) or omp_set_num_threads(N)). 

Without these the code runs 4 threads on a quad-core box. The problem
is: the extension module does not seem to honor the OMP_NUM_THREADS
environment var (it does for a "normal" Fortran executable). How do I
control the number of threads at runtime for the extension module?

best,
Steve