[Feature Request]: Data constraint choices shouldn't depend on the filtered data

[gogonzo]

Feature description

Data constrains currently depends on the filtered data. @npaszty suggested that constraints choices (None, Screening, Baseline) should stay unchanged no matter what happens in the data.

• Constraint range should still depend on a data. What to show in the case when range is NA?
• Please see keep_data_const_opts_updated

[npaszty]

just to clarify that the data constraint UI element should not have entries removed given right hand data panel filtering.
for example

• If user filters AVISIT on right hand filter panel to remove screening or baseline visit then the UI element should not react to remove these radio button choices from the data constraint UI element. They should remain static and any data constraint filtering created with them should remain active.
• If user filters SEX on right hand filter panel to remove "M" then it's great that the available data range in the data constraint text input reflects the range for the "F" population.

It's important to note that the data constraint filter filters out subjects from the analysis. So it's essentially a dynamic population filter vs. a static ones that are available in the ADSL dataset like SAFFL, MITTFL, ITTFL etc.

[npaszty]

@gogonzo @m7pr

all except correlation plot: see correlation plot related issue #287 . having both "Screening" and "Baseline" constraints displayed when there is no "Screening" visit in the study data creates a problem where if "Baseline" is selected to constrain on and then "Screening" is selected to constrain on the "Baseline" min/max value UI items a=continue to be displayed and if changed while "Screening" is selected will cause a validation message to be printed.