Sod shock L1 convergence test. This test is identical in setup and parameters
as the Sod shock test used to compare Shadowfax with SWIFT, but now is used
to explicitly check the convergence of the supposedly second order integration
scheme used by Shadowfax.

The test consists of a .xml and a .ini file, and a script that calculates an L1
norm that is used to quantize the convergence rate. The script assumes the
test is run 4 times, with 1,000, 8,000, 64,000, and 512,000 cells, and that
the respective snapshots are put in separate folders: sod1k_r, sod8k_r,
sod64k_r, and sod512k_r respectively. To run each test, generate the initial
condition:
SHADOWFAX_RUNDIR/icmaker3d --ncell XXX --setup sod3d.xml --filename ic_sod3d.hdf5
(you can generate a Cartesian setup by adding the --mode cart flag to this)
and then run the test:
SHADOWFAX_RUNDIR/shadowfax3d --params sod3d.ini

The L1_norm.py script then generates a file L1_sod_r.txt containing the 1D
equivalent resolution and L1 norm for every resolution, calculated as the
total difference between the final simulation result and the analytical solution
provided in the file rho.dat.