So that the program `eddy_graph` can read it, instead of deducing it
from the longitude coordinate.

We are trying to define clearly the directories and what Fortran files
they should contain:

- `Inst_eddies`: May only contain program units which make up the
program `inst_eddies`. If a program unit does not go into
`inst_eddies`, it should not be here. Not all program units going into
`inst_eddies` must be here: some could be in directory Common.

- `Inst_eddies/Tests`: Main programs of unit tests for procedures in
directory `Inst_eddies`. Plus procedures used only in these unit tests.

- `Overlap`: May only contain program units which make up the program
`eddy_graph`. If a program unit does not go into `eddy_graph`, it
should not be here. Not all program units going into `eddy_graph` must
be here: some could be in directory Common.

- `Overlap/Tests`: Main programs of unit tests for procedures in
directory `Overlap`. Plus procedures used only in these unit tests.

- Common: Procedures that are used in both programs `inst_eddies` and
`eddy_graph`.

- Common/Tests: Main programs of unit tests for procedures in
directory `Common`. Plus procedures used only in these unit
tests. Plus procedures used in unit tests of both `Inst_eddies` and
`Overlap` directories but not used in either program `inst_eddies` nor
`eddy_graph`.

Which read namelists. Motivations: diminish the number of arguments of
procedure overlap; lighten the main program unit. We conditionally
compile MPI calls in `read_grid` because we want to use variables of
module`read_grid_m` in procedure `read_snapshot`, and `read_snapshot`
is called from sequential program `test_nearby_extr`.