SRC: detect all-zero Y columns in GEDMD scaling

[nakatamaho]

This PR adds the missing all-zero column check to the JOBS='Y'
scaling path in the GEDMD routines.

In the existing implementation, the JOBS='S' / JOBS='C' path
normalizes the columns of X and counts zero columns while doing so.
If all columns of X are zero, the routine returns early with
INFO = -8, since X is the 8th argument.

The corresponding JOBS='Y' path normalizes the columns of Y, but it
did not count zero columns of Y and therefore did not return early
when all columns of Y were zero. This made the handling of the
SCCOLY path asymmetric with the existing SCCOLX path.

This PR updates the JOBS='Y' path to:

• initialize the zero-column counter before scanning Y
• increment the counter for each zero column of Y
• return early when all columns of Y are zero
• use INFO = -10, matching Y as the 10th argument
• document this condition in the routine header comments

Updated routines:

• SRC/sgedmd.f90
• SRC/dgedmd.f90
• SRC/cgedmd.f90
• SRC/zgedmd.f90

This is intended to make the input validation and scaling behavior of
the JOBS='Y' path consistent with the existing JOBS='S' / JOBS='C'
path. It also avoids continuing the computation after detecting that
the requested Y-based column scaling cannot be performed because all
columns of Y are zero.