[Optimization] `get_neighbours` can generate n samples and then validate them

[eddiebergman]

Just a note that get_neighbours could be optimized in a few places. Currently we seem to generate a number at a time and then perform valid checks and then transform, causing a lot of non-vectorized if checks.

This could be vectorized by generating n numbers, validating them together and then performing the transformation on all elements.

Just rough pseudo code

neighoubrs = np.array([])
while len(neighbours) < n:
    samples = rs.normal(value, std, size=n - len(neighbours))
    neighbours = np.append(neighbours, samples[<condition>])
    
if transform:
    neighbours = self._transform_vector(neighbours)
    
return neighbours

[eddiebergman]

A more detailed (untested) algorithm can be found here:

• PiBO Pull Request 2 - Beta Hyperparameters #222 (comment)

Some Cpython analysis should be done to improve and make sure this actually speeds up over the current implementation.

[eddiebergman]

This sampling was made significantly faster with a newer algorithm in #346. Implements some of the same ideas