On second thought, I find the default value confusing. We are
reverting commit b328ab6f.

This is required to read an edgelist coming from `eddy_graph`.

Since we now only write valid eddies to shapefiles.

Use out_cont%closed instead.

`Inst_eddies_12` uses the current revision and so does not contain
invalid eddies.

Move test `R4_15_splitted` before the tests that use it.

Instead of `Inst_eddies_10_splitted`.

Only useful for tests.

Because the numbering changed when we decided not to write invalid eddies.

I do not remember why I chose to write to standard error. First it is
clearer to write only error messages to standard error. Second the
order of messages on standard error coming from the Fortran code and
from the shapelib C library does not seem deterministic, and this is
annoying to compare runs.

Use test `Region_5_15_days` instead of outside experiment `Inst_eddies_4`.

We will need this new shapefile in program `eddy_graph` only for a non
longitude-latitude grid. We add a new dummy argument `grid_lon_lat` to
procedure `shpc_create` to decide whether we create the shapefile
`extr_proj`.

It seems simpler than putting it in `$large_input_dir`. I do not see a
drawback.

Use relative path to Overlap from `Inst_eddies/Tests`. For symmetry
with what we do in `Overlap/Tests`. So now the directories Overlap and
`Inst_eddies` can be added from the top-level `CMakeLists.txt` in
either order.

Add executable in the directoy containing the main program unit. This
is clearer as the executable file will be created in the corresponding
binary directory. Since the directory Overlap is added after
`Inst_eddies`, targets defined in `Overlap/Tests` are not yet defined
when processing `Inst_eddies`, so we cannot use `target_sources` for
these targets in `Inst_eddies`. So we use a relative path to
`Inst_eddies` from `Overlap/Tests`.

The program `test_weight` no longer exists, since commit 9ff8a9bb.

Introducing `config_graph_nml`.