Bug in 4.2.0 checkpointing code

[Jean-Noël Grad]

Dear ESPResSo users,

It has come to our attention that the checkpointing mechanism was broken 
in release 4.2.0. When reloading the state of the system from a 
checkpoint file, if the DIPOLES feature is part of the myconfig file, 
the magnetic dipole moment overwrites the particle director, which 
causes non-magnetic particles to be all aligned along the z-axis. This 
bug breaks momentum conservation and can lead to nonphysical forces 
being applied in simulations where the particle director plays a role 
(ex: relative virtual sites, Gay-Berne potential, anisotropic particles, 
active particles, etc.). Since the default myconfig file contains 
DIPOLES, most ESPResSo users are affected. In addition, when reloading 
from a checkpoint file, the particle angular velocity in the body frame 
is sometimes calculated using a default-constructed quaternion instead 
of the particle quaternion.

These issues have been fixed in the python branch of ESPResSo (PR #4637) 
and will be fixed in the upcoming 4.2.1 bugfix release (PR #4611).

If you are using 4.2.0 and need checkpointing for simulations with the 
ROTATION feature, you can update ESPResSo 4.2.0 locally with the 
following git commands:

```sh
git remote add upstream https://github.com/espressomd/espresso.git
git fetch upstream pull/4611/head:pr-4611
git cherry-pick 992f093ce947b
```

We are sorry for the inconvenience.

Best regards,
Jean-Noël Grad