Add flag to observations if required or not

[oyvindeide]

It is quite common for some observations to be missing, for example due to structural uncertainty, and ert should just deactivate these observations and move on, however there are other observations should not be missing, and ert should stop if these are missing. Today we treat all observations the same, and allow for all observations to be missing, and only stop data assimilation if there are no more active observations left.

One option is to allow the user to specify which observation should always be present, and ert will stop if they are.

[dafeda]

Mind elaborating a bit as to how observations can be missing due to structural uncertainty? What does it really mean that an observation is missing? Missing from what?

[oyvindeide]

A bit late to the party here, but a well could for example end up outside the grid, meaning that an rft-obseration could end up outside the grid. That probably means that realizations should be discarded.