LAPACKE interfaces and C prototypes for ?GEDMD(Q) do not match the implementation

[martin-frbg]

Description
While looking into a reported LTO type mismatch problem in OpenBLAS, I have found that the LAPACKE interfaces for the
Dynamic Mode Decomposition functions recently added in #736 appear to be completely bogus - probably being based on
an earlier draft of the implementation. (Missing JOBR and ZWORK/LZWORK parameters, wrong type declaration for WORK array, EIGS split into separate complex arrays for the real and imaginary component - even for the S,D forms).
In addition, the corresponding prototypes in the C headers lack the hidden arguments for LAPACK_STRLEN_END
Working on a PR
Checklist

• I've included a minimal example to reproduce the issue
• I'd be willing to make a PR to solve this issue