Add support for SG15 pseudopotentials

[mbercx]

The Yambo code requires a pw.x calculation as a starting point for the GW calculations. However, their code currently only supports Norm-conserving pseudos, and non-linear core corrections (included in the Pseudo-dojo pseudos) are not fully implemented. Hence, @mikibonacci has requested support for the SG15 pseudos, since he would like to use the protocol methods implemented in aiida-quantumespresso with this family of pseudopotentials.

A few topics of discussion:

• [ ] Do we keep extending the aiida-pseudo install command when providing support for new pseudopotential families?
• [ ] AFAIK, there is no detailed study that provides us with a set of recommended cutoffs for the SG15 pseudos. However, @mikibonacci suggested that perhaps the pseudo-dojo cutoffs might be a good guesstimate.

[sphuber]

If I understand correctly, SG15 is not a particular format, but actual pseudo potentials. The SSSP actually includes a number of SG15 pseudos. I think as a rule, we only consider adding an automated entry in aiida-pseudo install if the pseudo family is published, rigorously tested and provided through some persistent identifier, or at least a very stable location to download it from. People are free to create custom pseudopotential families. One can already create an SG15 family with recommended cutoffs with the current API and even CLI. All we have to do here is maybe improve the documentation on how to do this. But all the necessary functionality is there.

[mbercx]

I think as a rule, we only consider adding an automated entry in aiida-pseudo install if the pseudo family is published, rigorously tested and provided through some persistent identifier, or at least a very stable location to download it from.

Sounds reasonable. Does "rigorously tested" also mean that one or more stringencies with recommended cutoffs have been identified, as was done for the SSSP? Note that there are several publications on the SG15 pseudos:

http://www.quantum-simulation.org/potentials/sg15_oncv/

You can also download the full set in XML and UPF format from the site.

All we have to do here is maybe improve the documentation on how to do this.

Indeed, I suppose at some point we'll have to set up proper documentation on readthedocs 😅 - Would be happy to work on this, might be a good exercise for me to set this up from scratch. Will already prepare a set of instruction for @mikibonacci for the SG15 pseudos as a starting point.

[sphuber]

Does "rigorously tested" also mean that one or more stringencies with recommended cutoffs have been identified, as was done for the SSSP

What I mean here is that there is at least one set of cutoffs defined that has been tested with convergence tests which have been published, and this is not just "some set" that someone came up with off the cuff