Include update = false parameters in calculation of X_full in EnIF

[oyvindeide]

We are excluding parameters the user has specified that should not be updated in the calculation of x_full in EnIF, there are some reasons for that:

1. If the source of the parameter is DESIGN_MATRIX, we dont have control of the input parameters, meaning: a. They could be categorical, i.e. str labels b. They dont have a transformation, internally sampled parameters are 0,1 normally distributed, and are transformed into the user space, while parameters from the design matrix are directly in the user space
2. There could be scenarios where the user has parameters that are sampled by ert, but not updated

Should we treat 1. and 2. differently? Currently we are excluding both, but 2. seems ok to include, while 1. I am much more unsure about.

Related to: https://github.com/equinor/ert/issues/11511