Check output file suffix at the beginning of the script. No sense
going through the calculation if the suffix is wrong.

Do not modify dummy argument x of `plot_distr_funct`.

The old algorithm did not give correct results on tests
`Greece_trajectories`, `Test_order_edges` and `Traj_component`. We
abandon the idea of processing edges in date order. Instead, we now
process vertices in topological order.

This can happen if the edgelist file is not found.

`my_subgraphs` is only used inside the test on `"cost_function"`.

This can happen if the input graph file contains only an extracted
component.

Necessary because, with old input files, the graph did not have the
property `e_overestim`.

Instead of reading it from standard input.

Generalize `extract_component.py` so it can also write a graph in
graph-tool or graphml format. This is useful to extract a component of
the graph of segments including cost values.

Generalize `extract_component.py` so it can also read a graph in
graph-tool or graphml format.

This is useful for hacking.

The downside is that we store all duration values.

Test `args.min_duration` outside the loop on `expanded_traj`. Clearer.

This is not needed for further processing of the graph.

Run `trajectories.py` from a graph-tool file which does not contain
all the properties.

Add option `--debug` to write all properties to output files. The
properties which are not always written are not needed for further
processing of the graph. To implement the option, we define variables
for properties, so we revert in part commit 2e3c6606.

Instead of reading it from standard input.

Instead of reading it from a separate file.

The idea is to favor long trajectories.

Preparing for more complex script.