Frequency selection should be driven from filtered data table, not original dataset

[ChristianSpence]

Values in input$frequencies should be determined from the filtered df of selected variables/dimensions, not from the dataset df as currently. Will also need handling to remember previously selected values when contents of that filtered df change.

[ChristianSpence]

Current challenge with this is the filtering of user_datasets() into filtered_datasets() is done by code which is conditional on the presence of input$frequencies.

As we now have an all_frequencies factor object (ordered from daily to annually), we might be able to default the filter onto min(dates.freq) if input$frequencies hasn't yet been written to the UI, and then update the options after the first UI build