Do not show scale-variation uncertainties when NLO log grids are not present

[cschwan]

In this case the uncertainties are wrong.

[cschwan]

In general this isn't possible, because we can't reliably detect what LO and NLO is.

[felixhekhorn]

In general this isn't possible, because we can't reliably detect what LO and NLO is.

that doesn't seem correct: you can determine what NLO is: a order is NLO if and only if there exists an other order with exactly one lower power of coupling (either a_s or a_em)

[cschwan]

@felixhekhorn I was thinking of the possibility that one can store LOs and NLOs in separate grids, but after thinking about it that's a special case that we don't want to support. I'm going to change the CLI accordingly:

• if the grid is an FK table, then don't calculate scale variations
• if there are higher orders, starting at NLO there must at least be factorization-scale logs available, otherwise don't calculate scale variations (we might want to add a --force key to get the old behaviour)

[cschwan]

The CLI's convolute no longer show scale uncertainties, which is the job of uncert now. For this command the user must explicitly require scale uncertainties, and in this case we can assume they know what they're doing. I'm closing this Issue.