Consider parsing Warnings about "negative rho"

[ltalirz]

QE issues the following warning if the electron density becomes negative through the course of a calculation (see code)

negative rho (up, down):  5.559E-01 1.964E-01

This can happen e.g. in the case of PAW pseudopotentials.

• If this occurs at some intermediate points throughout the SCF, it can be an indication that SCF convergence may be difficult (see below). If the calculation does not converge, the workflow may e.g. decide to abstain from trying to resubmit it with different settings.
• If QE detects a case of very negative density in the final SCF step, we may want to raise a red flag here since it can put the validity of the result into question.

We found this to occur frequently with the SSSP for the lanthanides, and it often prevented the ensemble-DFT-based minimization procedure in sirius from converging.

Mentioning @mbercx for info

[mbercx]

Thanks for raising the issue, @ltalirz! I agree it makes sense to parse these and at least add a warning to the report. If the charge density is very negative, we could even consider adding an exit code to make the issue more visible.