Read `days_1950` and `eddy_index` from outermost and max-speed
shapefiles. When we only opened shapefiles in reading mode, we never
needed these fields. It was enough to read the date and eddy index
from the extremum shapefile. But when we open in appending mode, we
need all the fields.

Because now there is a case when we want to open in appending mode: in
program `inst_eddies`.

Note that programs `test_max_speed_contour_ssh` and `test_weight`
depend on fortranGIS through the module derived_types. Compilation
worked without linking to fortranGIS because objects used by
`test_max_speed_contour_ssh` and `test_weight` in
derived types do not use fortranGIS.

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

Thanks to commit 5ff9e1d2, we are reverting in part commits e1aa5bcc and
338dd393.

Thanks to commit 5ff9e1d2, we are reverting in part commit e1aa5bcc.

Than ks to commit 5ff9e1d2, we are reverting in part commits e1aa5bcc and
fda1e6cd.

Better to see a single label than several labels at the same position.

Instead of creating them. Motivation: avoid time spent concatenating
shapefiles in `inst_eddies.py`. This concatenation time, on Ciclad, is
of the same order of magnitude as the time spent by the Fortran
executable.

Motivation: we want to call `open_shpc` and `read_field_indices` from
program `inst_eddies` without making it an MPI program for one process only.

Because we are going to add preprocessor directives.

Change path from `Overlap/Analysis` to Analysis, following commit
3e41a613.

The program `test_write_overlap` was removed in commit c440ce20.

I added assert in commit 83123e30, I do not know why, it was not used.

`test_nearby_extr` does not need NetCDF since commit 83123e30.

Following commit e741203 of `GPC_F`. We have to add this
`find_package` of GPC in `Detection_eddies` because `GPC_F` now
declares a dependency on the abstract target GPC.

Remove `shpc_create` from use list in main program `test_overlap`. It
is no longer needed since commit 50d2f596.

Remove `shpc_create.f90` from source files of target `eddy_graph`. It
is no longer needed since commit 50d2f596.

Update to commit cb25161 from cmake repository.

Add test for existence of target `NetCDF_Fortran::NetCDF_Fortran`. If
the target already exists, because `find_package(NetCDF_Fortran)` has
already been called then we cannot create again the target. This
happens if we are a sub-project of a project that first creates the
librairies.

In case we are a sub-project.

Assuming the englobing project builds the corresponding librairies.

The procedure does not exist any longer.

A node is no longer identified by date and eddy index.

One can manually add the key cyclones to the returned dictionary.

Replace `tests_old_dir` by PWD for tests in the same series.

Move content of `tests.json` to the end `Tests/short_tests.json`.

There were only three files in `Overlap/Analysis`.