Add multiprocessing to parallelize stochastic background response computation

To compute the response StochasticBackground we currently sequentially iterate over all sky locations, get or create the associated point source, compute their 6 link responses, and add them to the 6 total link responses.

We could speed up this part by using a thread pool to compute each point source's link responses in parallel. Depending on the number of cores available, we could gain substantial gains -- though we have to be careful about the memory usage here.

The standard library's multiprocessing module seems to offer everything that we need. However, we need to move the response computation to pure functions (instead of class method) to easily spawn child processes without handling complex inter-thread communication.