Avoiding multiplication by 1000 for each instantaneous eddy.

Preparing to keep the radius in km and integrate a factor 1000 in
`FACTOR`.

To avoid repeated computation.

Rename function `calculate_radii_rossby` to `avg_radii_rossby`.

Take three lists as arguments of function `calculate_radii_rossby`
instead of a list of dictionaries, since this is what we have on the
calling side as vertex properties. Note that we no longer have to
distinguish the case where `ip_beg <= ip_end` in the main part of the
script.

Return a tuple of lists from function `node_to_prop`, instead of a
list of dictionaries, since this is what we need to store as
properties of a vertex.

Use vertex properties.

Preparing to compute statistics on the variation of these properties.

In method `__init__` of class `SHPC_class`. As a consequence, we
remove the dummy argument `def_orient` of `__init__`.

We want to be able to re-compute the cost, starting from the output of
a previous run of `cost_function.py`, when we have discarded the input
file of the previous run.

The old behaviour: averaging over a maximum number of instantaneous
eddies, becomes an option.

We do not use bisect because `max_delta` should be small, around 6,
and most often there should be only one time step between
instantaneous eddies, so the first guess at distance `max_delta`
should be close to the result. Moreover, we would need to define a
date function if we used bisect.

Preparing for generalization of these functions.

This does not change any result, it is just an equivalent
formulation. We are preparing for a generalization of the choice of
`ip_beg` and `ip_end`.

`comp_ishape` can now compute `i_slice`.

Compute `i_slice` in `comp_ishape` if `i_slice` is None. Add default
value of None for dummy argument `i_slice`.

Move dummy argument `i_slice` of function `comp_ishape` to last
position. Because we are going to add a default value.

These files will usually be created in the same directory.

Correct the test to know whether we can use part of the properties
already calculated for the beginning of the segment. When `len_seg ==
2 * n_days_avg`, `g.vp.inst_eddies[n][-n_days_avg:] =
g.vp.inst_eddies[n][n_days_avg:]` so
`g.vp.inst_eddies[n][:n_days_avg]` and
`g.vp.inst_eddies[n][-n_days_avg:]` are disjoint. The bug comes from
commit bdd0ce41.

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`.