Abandon the radius4 approach, because the resulting max-speed contours
had speeds significantly lower than those from Matlab and because the
approach with radius4 is difficult to extend to a non
longitude-latitude grid.

Demote dummy arguments nlon and nlat of `input_ssh` to local
variables. nlon and nlat are no longer used in main program unit
`inst_eddies` since commit 03033711.

Fuse the loops on outermost contours and maximum speed contours.

Because we want to fuse the loops on outermost contours and maximum
speed contours.

Rename file `set_all_outerm.f90` to `set_all_contours.f90`.

Rename procedure `set_all_outerm` to `set_all_contours` and module
`set_all_outerm_m` to `set_all_contours_m`.

Because `test_set_all_outerm` is now essentially the same as
`inst_eddies`. Replace `test_set_all_outerm` by `inst_eddies` in
tests.

Since this information is not in the SHPC.

Speed components are now defined in `set_all_outerm`.

Move computation of max-speed contours to `set_all_outerm`. This
requires adding dummy arguments u and v to `set_all_outerm`. The
motivation is to compute the max-speed contour for each eddy
immediately after computing its outermost contour.

Do not set components for extrema without outermost contours. We do
not care about these elements of the snapshot, they will not be
written to the shapefiles.

Restrict field to outermost contour instead of using `outside_points`,
following removal of argument `outside_points` of procedure
`set_max_speed`.

Remove argument `outside_points` of `set_max_speed`.

Move the second test on `self._ishape_last[i_slice][orientation]`
inside the first test. Avoid the double test on
`self._ishape_last[i_slice][orientation]` when
`self._ishape_last[i_slice][orientation]` is not None on entry of the
function. If `self._ishape_last[i_slice][orientation]` is not None on
entry of the function then, necessarily, the assertion on
`self._ishape_last[i_slice][orientation].size` and the definition of
`self._e_max[i_slice][orientation]` have already been done because
`self._ishape_last[i_slice][orientation]` and
`self._e_max[i_slice][orientation]` are only defined from the function
get_ishape_last.

When we look for the maximum-speed contour, we do not need to give
outside points. If the bounding box for the search is the bounding box
of the outermost contour and if the value of SSH for the contour C we
search is constrained to be between the SSH of the extremum and the
SSH of the outermost contour then C must be inside the outermost
contour. So it cannot contain any extremum that is outside the
outermost contour.

Aside from removing an un-necessary computation, the real motivation
for this commit is that, now, we do not have to look for all the
outermost contours before looking for the maximum-speed contours.

I have tested that the results are exactly identical on all
in-repository tests and on one year of the PhD_Lax domain.

Because it is not created by `inst_eddies_v6.py` in the test
`Mat_v6_to_shapefiles` but it is created in
`$tests_old_dir/Mat_v6_to_shapefiles` by `plot_eddy_contours.py` in
the test `Plot_eddy_contours_Mat`.

It seems that Matlab does not work on spirit without this option.

The advantage is that if `inst_eddies_Aviso.py` is modified then cmake
will be run automatically.

Useless since commit 38374ff0.