add NaN checks for input matrix A in GEEV, return error if NaN are in matrix A, NaNs in input are detected with LANGE (needed for scaling)

[langou]

This adds a NaN check on the entries of the input matrix A in GEEV,

GEEV would now return error INFO = -4 if there are NaNs in matrix A in input.

NaNs in input are detected with LANGE. Since LANGE is called for scaling anyway, there is no extra cost.

This follows @angsch's suggestion in issue #1128. This uses DISNAN instead of DISFINITE (as initially suggested). I am not opposed to create a DISFINITE function and use it, but I think it needs more decisions and consensus.

See as well PR#471 for a "similar" behavior.

[TiborGY]

Perhaps it would be worth noting in the documentation that these subroutines check for NaNs in A.
E.g. something like this:

*> \param[in,out] A
*> \verbatim
*>          A is DOUBLE PRECISION array, dimension (LDA,N)
*>          On entry, the N-by-N matrix A, which must not contain any NaN elements.
*>          On exit, A has been overwritten.
*>          Note: A is checked for NaNs on entry, and an error exit is performed if any are found.
*> \endverbatim

[TiborGY]

The diff has a lot of formatting changes. Perhaps it would be a good idea to split this PR into a "meaningful changes" PR and a pure formatting changes PR?

[langou]

Thanks for the feedback. In a perfect world, I agree with your comment. Let us merge and I'll try to be better next time.