Move script `extract_component.py` from directory Trajectories to
directory Overlap. It does not deal with trajectories and it does not
need the input graph to be a graph a segments.

Downgrade `sort_ssh` from argument of procedure `complete_ssh` to
local variable. `sort_ssh` is not used any longer in procedure
`set_max_speed` since commit 4a64f87d.

Instead of using a submodule. See Scott (2023 k1077, § 39.6.2 "Sharing
CMake Modules And Commands"). The drawbacks are: the directory cmake
is duplicated in each build directory; configuration requires a
network connection. The advantages are: it is easier to download the
project without git; we can more easily package a release.

So we do not call `time.perf_counter()` a second time after a simple
assignment.

This is useful to find the shortest path between the beginning and the
end of a trajectory, or to create a histogram of the number of
trajectories per component.

Bug fix: correct sign of variation of Rossby number and radius. We
have to be consistent with the computation of `delta_ro_mean` and
`delta_r_mean`. `delta_ro_mean` and `delta_r_mean` were computed by
subtracting origin from tail.

It does not make sense here to assume an ellipsoid: the computation of
the distance with $\cos \phi \delta \lambda$ and $\delta \phi$ is a
rougher approximation than treating the Earth as a sphere.

We avoid multiple conversions to rad of the same value, in the loop on
edges, since a segment can be head or tail of several edges.

Move functions `node_to_date_eddy` and `date_eddy_to_node` from script
`report_graph.py` to script `util_eddies.py`. `util_eddies.py`
contains more basic functions, it only imports
shapefile. `report_graph.py` imports networkx. It was not convenient
nor meaningful to require networkx for scripts that use
`node_to_date_eddy` or `date_eddy_to_node`.

Useless: there is a symlink to `util_eddies.py` in directory Overlap
since commit 56ca8ea3.

Read orientation from `expanded_traj.json`.

Installation has been tested with CMake `3.27`. Adding the maximum
version rids us of a warning message concerning policy CMP0144.

This seems to be a little faster, even if we do not use out-degree for
all the nodes. For `Global_1993_2023/Graph_anti`, this commit reduces the time
to define `traj_prop` from 23 s to 20 s.

We know that we are going to need in-degree for every node of the
graph. For `Global_1993_2023/Graph_anti`, this commit reduces the time
to define `traj_prop` from 31 s to 25 s.

This looked good but was much longer. For example, for
`Global_1993_2023/Graph_anti`, this commit reduces elapsed time
approximately from 24 mn to 2 mn.

This makes quite a difference in file size for a large domain. For
example, the size of `Global_1993_2023/Graph_anti/expanded_traj.json`
is reduced from 0.6 to 0.3 GiB. We do not need pretty display for
these files.

This is convenient for scripts `plot_traj.py`, `filter_traj.py` and
for hacking with function `get_duration` of `plot_traj.py`.