mdlc (dipolar layer correction) test needs to be checked

[RudolfWeeber]

The fact that the test did not fail in spite of the bug removed in #1568 suggests that the test didn't test the method sufficiently. The test data uses sqrt(3) as dipole moment, so the unintended cast to int in the calculation of the total dipole moment should have made a difference.

[jngrad]

Also the test (dipolar_mdlc_p3m_scafacos_p2nfft.py) could be modernized:

• remove print statements conveying the same information as one of the assertions
• use float(np.genfromtxt()) instead of python2-style file open
• deduce n_particle from the size of the particle array loaded from file
• instantiate all particles at once with numpy slicing
• maybe replace the duplicated code logic for err_* variables by a class static method to calculate the standard error or the mean from two input variables (the sample size can be deduced from those inputs, no need to involve data.shape[0])

[jngrad]

The proposed improvements to dipolar_mdlc_p3m_scafacos_p2nfft.py have been applied in 5ddc3f0.