Remove intermediary variables `avg_rad` and `avg_Rossby`.

We extract part of function `calculate_radii_rossby` into new function
`node_to_prop`. So we now separate reading data from the shapefiles
and computing the average radius and Rossby number. Thus, we prepare
for rewriting the algorithm in the main part of the script. The
quantity of reading is still the same, for now.

Rename variable `num_of_days_to_avg` to `n_days_avg`.

This allows readability of the property in graphml format. As vector
of double is more specific than python object, I suspect this commit
also makes the code more efficient. See also commit c7f7fd46.

Return a tuple from function `calculate_radii_rossby` instead of a
dictionary.

The vertex property `avg_Rossby` was never set after `n_valid_Rossby ==
0`. It appears that it had then a default value of 0. Better to
explicitly set it to 0.

Rename radii and rossby to `avg_rad` and `avg_Rossby`.

Rename `first_SHPC` and `last_SHPC` to `i_SHPC`. Rename `first_loc`
and `last_loc` to ishape.

This will allow to see repeated code and use fewer variables.

Return tuple or scalar from function `node_to_date_eddy`, depending on
new argument `only_date`, instead of returning a dictionary. The
function is easier to use. Also, it saves a useless computation if we
are only interested in the date index. Returning a dictionary was more
justified before commit 43d2eff6, we we also returned the date.

Simpler, no subtraction.

Return from function `calculate_radii_rossby` the average radius and
Rossby number instead of the sum of average radii and the sum of
Rossby numbers.

Since the length of the segment is lower than or qual to
`num_of_days_to_avg`.

Pass a restricted list of eddies to `calculate_radii_rossby`: the list
over which to calculate mean Rossby and radius. Instead of passing the
whole segment and the indices of eddies in the segment.

Preparing to introduce an option.