Integrating work done by Mevludin Isic. We add the possibility to
split `id_child` because it takes more than 2 GiB for the global
domain and thus we cannot save the whole variable to Matlab v6
format. We split `id_child` along the time dimension. As the Matlab
code for this is the same for anticylonic and cyclonic eddies, we
extract it from the script `overlap.m` to a new script
`id_child.m`.

So we can easily import the functions in the file.

Instead of filenames constructed from the orientation. So we need to
create symlinks before running the script. It is clearer this way that
the script processes a single orientation, and that the processing
does not depend on the orientation. The script does not take a command
line argument any longer.

Rename variable `N(anti|cyclo)` to `N_eddies` in file
`N(anti|cyclo).mat`. Motivation: allow more transparent processing in
`overlap_v6.py`.

Only place where it is used. So we are reverting in part commit
68da90cd. We use again the function `days_1950` in script
`overlap_v6.py` since commit f46e1bc8.

The node identifiers in Matlab use the date index in the array of
processed dates. This is not simply the difference between the current
date and the first date if there are missing dates. This means that
converting between node index in Matlab and date and eddy indices, one
needs the array of processed dates. Morever, the conversion from date
and eddy indices to node index requires a search through the array of
dates. So we do not keep the Matlab node identifier. We use k and n as
described in the documentation.

For this commit, we need to revert in part commit 68da90cd. We have to
read `date_num` from the Matlab files, and we create function
`adjust_n` to convert from Matlab node identifier to the new node identifier.

Because we want to be able to easily read the edgelist with several
graph programs and they would not all accept the same
header. `read_edgelist` from Networkx accepts a comment line starting
with a special character. Not very elegant to put column headers after
a #. Anyway, the headers predecessor and successor are obvious for an
edge list.

Clearer. `e_max` is now the number of eddies of the current
orientation at the current date. See documentation.

Write node index instead of date and eddy index to the edgelist
file. Motivations: stay closer to Matlab, at least for now; some graph
processing programs other than NetworkX will not easily read nodes
that are tuples. Consequences of this change are that we no longer
need `date_num` from the Matlab files and we no longer need the
functions `date_index` and `convert_id` in the script
`overlap_v6.py`. So move these functions for now to a new file
`convert_id.py`.

n is for node number, e for eddy number at a given date.

It is not really meaningful to keep function `print_edgelist` since
commit 1e4d2ee1.

e1 is directly i1 + 1 and we just need the new function `date_index`
to process k1. Additional benefit: we process k1 out of the loop on
`id_child_1`.

Which is separate for cyclones and anticyclones. For now. Then we do
not need the argument cyclone of functions `convert_id` and
`print_edgelist`, and we do not need to read `Nanti.mat` for
processing cyclones.

`inst_eddies`: instantaneous eddies. `overlap`: overlapping.

The choice is made with a command line argument. We need this fix
to compare with results of program `eddy_graph`. This also makes the
script use less main memory.

`sio.loadmat` creates an array of dtype short int. This can produce
integer overflow in function `convert_id` when adding `Nanti[k]` and
`eddy_index_Matlab`.

And print to `edgelist.csv` instead of standard output. Because we
want to print other things to standard output.

We need to manipulate the smallest possible files when we deal with larger
spatio-temporal domains.