Run code generation in parallel

[mbanck]

As far as I can tell, stage1 (generator build) and stage3 (compilation) are done in parallel via make -j4 or whatever. However, stage2 (code generation via build_libint) seems to be serial. This is not a problem for smaller MAX_AM, but once you go higher, it can take a long time and it sounds like something that should be possible to run in parallel in principle.