[Draft] Move memory transpose logic to the Fortran client

[mellis13]

This PR moves the memory transpose logic from the C++ client to the Fortran client. For large messages originating in the Fortran client, it was demonstrated that intrinsic Fortran methods are much more efficient at doing the transform. Going forward, the C++ client will only be responsible for c-style contiguous and nested memory layouts. This greatly simplifies the internal logic of the clients.