SP with global inhibition should always obey numActiveColumnsPerInhArea

[mrcslws]

See discussion: numenta/nupic.core-legacy#1164 (comment)

With local inhibition, the inhibitionRadius varies with the statistics of the input, so if you're using the numActiveColumnsPerInhArea parameter, for certain inputs you might end up with way more active columns than you want, because the inhibition areas might be small. So it's probably better to use the localAreaDensity when using local inhibition. Anyway, the code does some work to minimize this effect: it makes sure to use a density no greater than 0.5. https://github.com/numenta/nupic/blob/fbd7ac40a23a536317c76b85089e90f5bc5f1fe2/src/nupic/research/spatial_pooler.py#L1400

This logic also happens in the boosting code.

So this check is addressing the problem that numActiveColumnsPerInhArea is awkward with local inhibition. We should move this check into the local inhibition code path. The global inhibition path should simply obey the parameters, since the inhibitionRadius is predictable and doesn't change with the statistics of the input.