This removes OpenMP-based parallelisations that are present originating
from the AIX-based versions. These are removed to make place for an
updated, hybrid MPI & OpenMP approach.

Considers files:

- CALAVBOLD.for
- CALEXP2T0.for
- CALPUV2C.for

This fix resolves the patch applied in OpenDA that comments out these
print statements, specifically `CGATEFLX.patch` and `RWQC1.patch`.

A print is inserted before invoking `READ(*,*)` to notify the user of
the suspended state of the application. This is similar to the message
emitted by `PAUSE`. The behaviour is now slightly different:

- To continue: ENTER should be provided in stead of 'go';
- To terminate: a signal needs to be sent manually.

This resolves the warnings raised by the compiler due to lines
containing unused jump labels. For lines that only contain format
specifiers or CONTINUE statements, the lines are commented. For lines
that contain other instructions, such as calls to READ, only the label
is removed.

The non-standard function SECNDS is replaced by SECOND throughout the
source code. Since the alternative SECOND does not allow to provide a
reference time, the time difference is obtained by subtracting the
original reference value instead.

For details on the non-standard call:
https://gcc.gnu.org/onlinedocs/gcc-4.9.4/gfortran/SECNDS.html

This makes the implicit while loop for the READ statements explicit by
moving the READ statement into a `DO WHILE` statement that runs until a
satisfactory read is performed.

This does not change the functionality, it merely makes the loop
explicit and thereby suppresses the compiler warnings.

All implicit conversion is made explicit by inserting the required type
casts to match the assigned variables.

This resolves the corresponding warnings raised by
`-Wmaybe-uninitialized`.

This reverts commit 3dceaaf.

Some of these are old loop variables that were not removed when the
corresponding parallel loops were stripped out. Others are simply
additional unused variables.

These are commented here and seem unused.

The filenames '${NAME}_mpi.for' implement the MPI counter parts of the
subroutines present in '${NAME}.for'.

In NIER these variables are defined in Var_Global_Mod and not explicitly
passed towards the subroutines. The subroutine signature is modified
accordingly.

- Replace gfortran with an MPI aware compiler (mpifort)
- Add targets for the newly added MPI module and MPI sources
- Add compiler flag -fallow-argument-mismatch to suppress errors
  resulting from type mismatch in MPI library calls.

Includes a note regarding missing patches in the NIER version of
source file CGATEFLX.

* Add missing Green algae salinity tox loops
* Converts .AND. to .OR. to be consistent with all other comparisons
* Modify if-statement IWQFCB to check with zero

7b484cc

This introduces the MPI patches present in the original Fortran source
files, such as including the MPI module and various occurances of
checking the current rank through MYRANK, mostly to guard read, write,
or debug statements.

This makes the number of threads extracted consistent with the OpenDA
wrapper's approach.

This propagates the MPI patches into the OpenDA wrapper. The MPI
environment is initialised when the wrapper is set up. The domain
decomposition seems required before the second model init, as is done in
the EFDC.for main file. When the wrapper is destroyed, the MPI
environment is destroyed as well.

NOTE: This *ONLY* considers running with at most 1 rank and no care is
taken to restrict certain operations to only the first rank, as would be
needed for file IOa and/or log printing.