Investigate optimal value of tau0 for NHC thermostat

[danielhollas]

The current value of the tau0 is 0.001 ps, which is very very low, likely tailored towards the PIMD application (but even there it might be suboptimal). I think for most other MD programs such as Amber or Gromacs the value is more like 1 ps. But it is true they are tailored more towards simulations on nanoseconds (and beyond) timescales, whereas in out ab initio world we're typically reaching tens of picosecond. Shorter value guarantees that the system is equilibrated quickly towards target tempertature, but it might hinder sampling of slow modes / process such as difussion / barrier crossing. We should think about a good test case to verify this and choose a perhaps more reasonable value as the default, at least for classical MD. The other option would be to have no default and leave the choice to the user, but we should still do some testing and provides guidance.