For programs `inst_eddies` and `test_get_1_outerm`, following commits
9a59d13 and 9dc139d.

Following commit b0e3ce55.

Now that `inst_eddies` reads `main_nml` only on the first date, we do
not need to create a file `main_nml.txt` before running
`inst_eddies.py`. `inst_eddies.py` can just read `main_nml` from
standard input.

Thanks to commit 5ff9e1d2, we are reverting in part commit f04aea8b.

Replace `SHPC_(anti|cyclo)_all_dates` by `SHPC_(anti|cyclo)`.

Thanks to commit 5ff9e1d2, we are reverting in part commit 83123e30.

Replace positional argument template by optional argument. This
argument is now needed only in case we process several dates, since
the first date is now read from the NetCDF file.

Even for a run with a single MPI process, otherwise execution will not
work on some systems (jean-zay).

Replace `/proju/lmd-oce/data/AVISO/DT_2018/2006/01` by
`$large_input_dir`.

Since `Inst_eddies_loop` is in the same series of tests as
`Nearby_extr`, take it from `$PWD` rather than `$tests_old_dir`.

Make it 15 days long because we want to use it in place of
`Overlap/Tests/Input/Region_2_SHPC`.

Instead of reading `extr_map.nc`, `ishape_last.txt` and shapefile
extremum. The problem was the consistency between `extr_map.nc` and
the two other files. `extr_map.nc` contains eddies with the two
orientations while extremum now contains a single orientation, with
eddies renumbered.

This SHPC was made with an old version of `inst_eddies`: it contains
cyclones and anticyclones output together. We use
`$tests_old_dir/Inst_eddies_loop` instead. See also commit 8ed324ff.

Move the test `Inst_eddies_lopp` earlier in the test series because we
are going to use its results in subsequent tests.

Run it with two dates and the new input NetCDF files. Two dates so it
can replace the SHPC that was in `$src_dir/Tests/Eddies_region_4`.

Since it is a test on script which is in Analysis.

Use higher level procedures in main program
`test_set_max_speed`. Allows to output easily more information. Then,
for example in `test_mean_speed.py`, we will be able to read extremum
position from the shapefiles instead of asking the user to define it.

Instead of asking the user to define it. Motivation: make
`test_set_max_speed` easier to use.

Instead of creating symbolic links to the shapefiles. Clearer and
easier to use.

Use higher level procedures in main program
`test_set_max_speed`. Allows to read easily more information: cyclone,
`ssh_extremum`, `coord_extr` from the shapefiles. So we do not ask
this information to the user. The drawback is that we have to call
MPI (because of `read_field_indices`) even if we do not really use
it. We move the calls to `find_package` for EZMPI and `GPC_F` to the
top-level `CMakeLists.txt` since these libraries are now necessary in
both `Inst_eddies` and Overlap.

Use higher level procedures in main program
`test_get_1_outerm`. Allows to output easily more information. Then in
program `test_set_max_speed` for example, we will be able to read information
about the extremum from the shapefiles instead of asking it to the
user.

Create separate collections of shapefiles for cyclones and anticyclones.

Preparing for two collections of shapefiles. We create the
procedure `write_nml` so we will be able to call it twice. But it
needs to access the namelist. So the namelist must be at the module level.

The non-interactive test requires adding a save option to the script.

This is more robust: the default value of `max_radius_deg` remains
reasonable when we use input files with different resolutions.

Homogenize with other dates, preparing for a loop.

Revert parts of commits 845752bc and ef7abf1b. It is true that having
the input files in the repository allowed us to run tests for the
first time more easily. But the drawbacks were:

- The large number of input files obscured the repository tree.

- Some files are input files for a test and are actually produced by
another test. These files did not evolve automatically with the
executables that produced them.

- It was a burden to log how these input files were produced.

To alleviate the problem of running the tests for the first time, we
take the input files from tests in the current series of tests, when
possible, instead of from the old series. (Before commits 845752bc and
ef7abf1b, the input files came from `$test_old_dir`.) We do not always
do this because we want to be able to run the various series of tests
independently. For example, we want to be able to run the Overlap
series before the `Inst_eddies` series. So, for the files that are
produced in one series of tests and used as input files in another
series of tests, we go back to pre-commit ef7abf1b and use
`$test_old_dir`.

Take into account new structure of repository in JSon test files.

Also split the tests, Fortran source files, Python files and JSon
files. Keep at the top level the files used in both `Inst_eddies` and
Overlap.

Motivations for the split:

- The top directory contained many files.

- It may be useful to compile only one of the two sub-projects. For
example, only `Inst_eddies` on Ciclad, where MPI 3 is not available.