P3M-based electrostatics and magnetostatics tuning functions ignore errors

[jngrad]

The electrostatics tuning function runs into an infinite loop if the system time step is not set. Magnetostatics tuning is not affected. Regression introduced in 4.1.0 by #3141. MWE:

import espressomd
from espressomd import electrostatics
system = espressomd.System(box_l=3*[1])
system.part.add(pos=[[0, 0, 0], [.5, .5, .5]], q=[-1, 1])
# system.time_step = 0.01
solver = electrostatics.P3M(prefactor=2, accuracy=1e-2)
system.actors.add(solver)  # infinite loop

[jngrad]

There are more issues with electrostatics tuning: several errors are simply appended to the standard log string, preventing handle_errors() from halting the flow of the program and leaving the system in an undefined state. Magnetostatics tuning is also affected.

[jngrad]

The magnetostatics dp3m_rtbisection() function doesn't throw on error, and the error code it returns isn't checked, so we end up in an infinite loop...

[jngrad]

The script interface for P3M in the metallic case is broken since before 4.0.

[jngrad]

The code samples in the user guide for electrostatics/magnetostatics don't run in pypresso. The user guide says P3M requires a Bjerrum length when it actually requires a prefactor since 4.1. The user guide says using DipolarP3M requires reading the P3M literature but doesn't provide the list of P3M citations.

[jngrad]

The epsilon parameter is changed to 0 during tuning in P3M and P3MGPU classes since ESPResSo 4.0.0. DipolarP3M is not affected.