PCA min and max wrongly determined

[LucaMantani]

When doing a pca fit, the max and the min are obtained by taking the vector of min and max values and rotating it (see line of code). However, this is not a correct procedure.

This can be seen by imagining that we are fitting:
c1: {max: 1, min: -1}
c2: {max: 1, min: -1}

Imagine that a direction is now PC0 = c1 - c2.
The min of PC0 is obtained when c1 = -1 and c2 = 1, i.e. when one is at the max and the other is at the min of their range, not when they are both at the minimum value.