#164 columns.R should not use type "identity" in dependencies

[Michele Gubian]

Example:

from GUI load IA, run ldt on a stimulus set, create a column.R on (ldt default_observer).
Then run a parameter exploration, which will add a new column to (ldt default_observer). showning the parameter value that was explored.
Here's the problem: the settings widget of the column.R dataframe_view that depens on (ldt default_observer). does not show the new column until you switch to another df and back. This is because we are using "identity" on df, rather than the get_description_element thing.
I am not sure whether there was a reason to keep it this way or not.

[James S. Adelman]

Currently only dataframe_for_activities (i.e., layer observers) have associated colnames and colname_dimensions objects.

So there's no description that contains the column names.

So lazyNut asks the GUI to be conservative and re-check the dataframe column names via command if the dataframe's value has changed in any way.