Force process grouping in theory covariance module

[voisey]

As discussed here #651 (comment), in the theory covariance module we should force metadata_group: nnpdf31_process or at least set it as a default (maybe this is better in case someone wanted to change it down the line? Although I'm not sure if there's an obvious case when anyone would want to do that). We want to do this because otherwise the default behaviour will be to group data by experiment in constructing the theory covariance matrix, which would lead to incorrect results for all use cases we have so far, unless you manually set metadata_group: nnpdf31_process in the runcard. Note that this is waiting for #651 to be merged, but that should happen soon.

[voisey]

@RosalynLP Is this at all related to other things that are happening with the data keyword? If so, perhaps we should take care of this now?

[RosalynLP]

@voisey I don't think it's related, it's a separate issue, it's more of a safeguard in case people don't add the line manually

[voisey]

Hey @RosalynLP, looking at #1110 it seems important that this is addressed soon. Do you agree with the sentiment of this PR?

[RosalynLP]

Yeah needs doing soon, I agree! Will prioritise