Possible bug in KKT equilibration

[goulart-paul]

A possible bug in the equilibration procedure. I am not certain that this is a bug because it depends on whether the P matrix can be only upper (or lower) triangular when equilibrating. The problem is this:

• Suppose that P == triu(P) inside the ruiz scaling procedure.
• The functions calls kkt_col_norms! here.
• That function in turn calls col_norms! here, where the col_norms! function itself does not seem to check if a matrix is represented only by its upper or lower triangle.

if this is a real problem then I can fix with a PR since I fixed it already in a different package. Whether it is necessary depends on whether P is guaranteed to be fully symmetric or not.