Parallelize q-distance computation

[JinLi711]

Parallelize the q-distance computation to use multiple cores. Make sure that the overhead is not large. We can try to put every other row in the q-distance matrix to one core. For example, if we have two cores, then we can put rows 0, 2, 4, etc to CPU 0 and rows 1, 3, 5, etc to CPU 1.

[JinLi711]

However, it's not a good idea to use multiple processes because every time a new process is created, there is a huge overhead in copying data over to multiple CPUs. Ideally, we would use multiple threads but this is not simple to do with python GILs. We can go around this issue by writing the code in native C code or use Numba but that would require a lot of code rewriting.