Merge the concept of parameter groups

[pinkwah]

ERT 2 has a concept of parameter groups, and parameters. The user specifies parameters in a file, such as COEFF_A UNIFORM 0 1. Then, they use the GEN_KW keyword to specify a parameter group, and give it a name. The "flattened" name of the parameter is COEFFS:COEFF_A if the group name is COEFFS. Different parameter groups can be sampled from the same parameter file.

ERT 3 has a similar concept, but calls the contents in the file "parameter variables". The record name corresponds to parameter group as above. Thus, a parameter coefficients has variables a, b, c, and when sampled is put into a record name uniform_coefficients. Thus, by ERT 2's scheme, we would flatten it into uniform_coefficients:a, uniform_coefficients:b, uniform_coefficients:c.

Since we probably want to use ERT 3's model, ERT 2 should combine parameters into parameter groups. Ie, it should send COEFFS, with the columns COEFF_A, COEFF_B, COEFF_C.

[mortalisk]

@DanSava Do you think this issue is solved with the work you recently did? Or do you have additional input?

[berland]

The ERT3 model has now shifted, as scalar parameters are now a thing (equinor/ert#2934).

Then the "parameter group" concept in ERT3 is just a parameter after equinor/ert#2934.

A parameter in ERT3 can contain "variables", which just means that its data can be indexed like a dictionary. A parameter group in ERT3 may in the future be closer to the ERT2 model (equinor/ert#3114).