Bug fix: extend `inst_eddies` of predecessor with `inst_edddies` of
circumvented node, instead of just name of circumvented node. The bug
is demonstrated in test `Edgelist_Isic_2` where segment 10 missed eddy
12 and segment 13 missed eddy 15.

Initialize the list `inst_eddies` with vertex name in the initial
graph, which is the graph of eddies. Else the list for a segment may
miss some instantaneous eddies. This bug is demonstrated in test
`Edgelist_Isic_2` where segment 10 missed eddy 10 and segment 29
missed 29.

It is useless to add to an attribute to v since, after the current
iteration, v will be isolated, and so it will not be used in remaining
iterations, and moreover it will be deleted after the loop.

Do not delete delete v just because it is isolated. v may have become
isolated at an earliest iteration because it is the first eddy in an
isolated segment. The bug is demonstrated with test `Edgelist_Isic_2`,
where vertex 29 was erroneously deleted.

So the comparison with `segments.py` is easier.

So the comparison with `segments.py` is easier.

So the comparison with `segments.py` is easier.

For easier comparisons.

Following creation of sub-directory `Trajectories/Tests`.

Same as `edgelist_Isic_1.csv` but with lines out of order.

Rename `edgelist_test.csv` to `edgelist_Isic_1.csv`.

May be useful for comparison with `segments.py`.

Accept input graph of segments without cost functions and accept
absence of file `traj_segm.json`.

The attribute written was always true.

Since we now only write valid eddies to shapefiles.

Otherwise we get a harder to understand error from numpy.

This is useful is an SHPC is missing an orientation.