Keep file versions and observation dates separate when reading a CSV

By The Sideline A data file can have a release date that differs from the date of its observations. A download time is a third timestamp. Treating them as interchangeable can make a dashboard appear fresher than its inputs. The Sideline's public five-row CSV sample provides a small example. It has sample_version and value_as_of fields, plus value_type to identify the modeled quantity. In the…