multiprocessing: from pool.apply to pool.imap_unordered

[apasto]

Here we do need the synchronous calls and ordered execution of pool.apply (pool is an instance of multiprocessing.Pool).
The reason why I used it was mainly out of habit, and ease of providing the arguments (same for all elements) in kwds:

decorrelate-grids/decorrelategrids/rollinglinreg.py

Lines 190 to 204 in c41f8ba

	pool = mp.Pool(n_processes)
	for element in valid_elements_idx:
	(
	out["c0"]["c0"][element[0], element[1]],
	out["c1"]["c1"][element[0], element[1]],
	out["rv"]["rv"][element[0], element[1]],
	out["pv"]["pv"][element[0], element[1]],
	out["ie"]["ie"][element[0], element[1]],
	out["se"]["se"][element[0], element[1]],
	out["np"]["np"][element[0], element[1]]) = pool.apply(
	rolling_linregress,
	kwds={
	'a': A.to_numpy(), 'b': B.to_numpy(), 'e_i': element,
	'hw_x_i': window_halfwidth_x_i,
	'hw_y_i': window_halfwidth_y_i})

Let's try to migrate to pool.imap_unordered and see how it turns out.