Drop `yamldb` as default

[giacomomagni]

We should drop the usage of yamldb as default, in favor of validphys once:

NNPDF/nnpdf#1942

is merged. But we can keep the possibility to use yamldb.

[alecandido]

I'd say that, for the sake of maintenance, there should be only one way to do things, to tame complexity.

If you need an easy way to patch it, that should be provided by vp (or the VP->Pineko interface).

[Radonirinaunimi]

I'd say that, for the sake of maintenance, there should be only one way to do things, to tame complexity.

If you need an easy way to patch it, that should be provided by vp (or the VP->Pineko interface).

But I think this is a big issue for external users. External users should not have to implement vp-commondata in order to just generate predictions. So still allowing for the yamldb would alleviate this.

[alecandido]

First, there are not many external users for Pineko, since its only goal is to deal with FkTables.

Then, I'd say that not all of the vp-commondata should be required for Pineko, so you could trim them to a minimal specification.

If you want to use Pineko w/ and w/o vp, the best would be that Pineko states an interface, and this interface (by chance :P) is also implemented by vp-commondata.

@scarlehoff

[scarlehoff]

@comane will be in charge of creating a commondata/theory package out of vp with only the parser so that it can be used for pineko, for the gaussian process, etc.

[alecandido]

Ok, but it will be the same one used in vp, isn't it?