terrain/steps.py

it appears we are taking one too many round trips to do when pre-fetching children. This can be very expensive as the tree gets wider the deeper we go. For example, in courseware we want depth=2 (course, chapter, sequential). But looking at log output we were also getting verticals, which there can be a lot of. This should cut down on the total data we are grabbing from the DB.

Includes corresponding documentation changes.

Per request from Harvard: LH #276

message when we are queued as well.

rather than using the login dialog.

if we're loading a course module with depth = 0, then we don't need to fetch/compute inherited metadata

- improve docstrings
- only pass in the state for a particular input and
not the whole dictionary
- refactor some common code
- minor syntax cleanup

world.css_click() helper instead of directly
calling splinter.