So user is redirected to the correct page after uploading data

When process_items is called without a request, redirect to data-view (don't try to double guess the tab)

Provide the tab instance to get_context_data, and use that to work out the URL to redirect forms to.

Remove outer html from the view and edit page, this is now taken care of by the tabbed page rendering.

Refactor the ViewsItems view, which covered the view and edit table, into a separate (non-view) class under tabs/view_and_edit_table.py. When ready, this will plug straight away into the new tabbed page functionality - in the mean time there is a place holder view to display the view where it used to be found.

Ensure we use the right taxonomy for the question category. The taxonomy in use is still hard-coded, as that is something we will look at later - this commit was only intended to restore functionality to the existing level.

There are two new item types 'question' and 'rumor'.
Item types for GeoPoll data are 'question'.
Item types for RapidPro data are 'rumor'

Allow the chart count widget to have a time period. When present, apply the time period and display a legend detailing what the time period is.

-- replaced {{name}} with {{title}} on chart widget template
-- added h2 tag around bar chart table template title
-- fixed icon span class

ViewItem.get_category_options should return the options as we want them - we shouldn't have to process them later. So make sure it returns (short question, short question) as we want both the value and the text to use the short version.

Refactor add_items_categories and add_categories into a single call, and make the taxonomy slug configurable at that level (it's now hard-coded in process_items). Update tests to match the change and ensure we can update multiple items at the same time.

Make the default empty batch command a line of dashes under the 'Actions' header. We can't really do better than that without js.

Re-introduce the per-row drop down to update each record's category as well as the mirrored buttons at the bottom of the page. This means that we have a single form that can either do a batch update, or an update of values set per row. The different action is determined by the submit button that was used - which is always named 'action'. Because we have a mirrored select drop down for the batch action, we need to know which one to take into account - so the value of the submit button indiciates which one was clicked, of the form <action>-<placement> (eg. save-top) and the name of mirrored inputs indidicate where they are located, of the form <name>-<placement> (eg. batchaction-bottom). The ViewItems class has a static method to find the active placement and rename the parameters accordingly. Some of the old code was re-introduced, though some of it was refactored (eg. what used to be get_categories, to fetch each row's selected category, has been moved to ItemTable to parallel get_selected, and is now named get_row_select_value (it's a more generic version, looking at the future when we'll update the region with the same mechanism.)

Now if a user attempts to log in from a login page with a stale
CSRF token, they will be redirected to the dashboard or, if they
have logged out, to a fresh login page.

Add tests for what happens when `add_categories` gets called with nonsense items and category names, should fail gracefully
https://trello.com/c/fWrvZDau/77-clear-out-technical-debt-from-working-on-taxonomies