Improve metadata processing

[VladimirShitov]

• In Stephenson dataset:
  • column Smoker has a value Not_known, should be NA
  • column Status_on_day_collection has levels with 1 and 2 observations, should be merged or set to NA (alternatively, this column can be dropped as there is Status_on_day_collection_summary)
  • column Outcome has level unknown, should be set to NA
• In COPD dataset:
  • Column Bronchodilator_Use has small categories, should be renamed or merged
• In HLCA dataset:
  • Column cause_of_death has a lot of small categories, should be grouped
  • Column lung_condition has some small categories
  • Column lung_condition has category Healthy (tumor adjacent), which should be renamed to Healthy with tumor information contained elsewhere
  • Column lung_condition contains COVID severity levels, might be better to group them together (done in disease) and move severity to a separate column
  • Column sequencing_platform has small categories and one mixture of two categories
  • Column smoking_status has a small category hist of marijuana use, which should probably be merged with active
  • Column assay has small categories
  • Column sex has category unknown, should be set to NA or predicted
  • Column self_reported_ethnicity has small categories, might make sense not to use it at all