Changing my mind. In an SHPC, store shapefiles in
"Slice_{slice}/{orientation}" instead of
"{orientation}/Slice_{slice}". Two reasons: easier to store accessory
files from running `inst_eddies`, such as `perf_report` and output
from Torque submission script; clearer that a slice should contain the
same dates for both orientations.

Read `used_nml.txt` at the SHPC level, only one file for both
orientations. Remove dummy argument exist of procedure config. Demote
`config_nml` to local scope. Remove procedure `write_nml`.

Remove procedure `write_aux`. Write to `ishape_last.txt` in procedure
`write_snapshot` and write `grid_nml` in procedure `input_ssh`. In
`input_ssh`, if `grid_nml.txt` already exists, check consistency with
input NetCDF files. Downgrade dummy arguments `corner_deg` and
`step_deg` of `input_ssh` to local variables. Read and write
`grid_nml` at SHPC level, only one file for both orientations and all
slices.

Assume that a slice is in a directory
"{shpc_dir}/{orientation}/Slice_{slice}" where shpc_dir can be any
name, orientation is "Anticyclones" or "Cyclones" and slice is an
integer. Do not read or create any longer a file "orientation.txt" in
the slice directory. Add procedure `get_slice_dir`. Add dummy
arguments cyclone and slice to procedure `shpc_open`. Add dummy
argument slice to procedure `shpc_create`.

Not needed since commit 53e5c983.

Not needed since commit 6dc504df.

Rename `main_nml` to `config_nml` in module `config_m`.

Extract code from program `inst_eddies` into new procedure `write_snapshot`.

Extract code from program `inst_eddies` into new procedure `input_ssh`.

Split derived type `shpc_slice_handler` into `shpc_slice_handler` and
`shpc_slice_meta out`. Do not mix what is necessary to open, create
and close, analogous to a logical unit, with what comes from reading
the files.

Create component `unit` for `ishape_last.txt` in derived type
`shpc_slice_handler`. Open, create and close `ishape_last.txt` at the
same time we open, create and close the shapefiles. So we use
`read_opcol` instead of `read_column` when we need to read
`ishape_last.txt`.

We are more clearly separating opening files and reading from them.

This is no longer needed since commit d16d3f1c.

Use `e_overestim.txt` instead of `node_id_param.json`. Since commit
43d2eff6, we wrote only `e_overestim` to the JSon file. Simpler to
write the value alone in a text file.

Add tests for a case where the speed is not defined on a whole segment.

Move tests `Plot_eddy_contours_Mat` and `Stats_Matlab` to
`Convert_Matlab/Tests/long_tests.json`. Thus, we group the tests that
depend on the availability of Matlab.

The modification of `sys.path` was added in commit 034d96fe to have
access to `util_eddies`. It is no longer useful, since commit
d16d3f1c, when a symbolic link was added.

To ease comparaison of runs. Also to unclutter standard output when we
run thousands of dates.

Because it contains timing.

Following update of `test_compare.py`.