The reset attempts web interface now defaults to only resetting the
current problem. Added "Reset all problems" check box to the reset
attempts form in the 'Proctor Student Admin' interface to enable
resetting *all* problems for a single user.

Recent updates to policy.json in 3.091r course xml changed the grade
data structure returned by edX from:

{'totaled_scores': {'Assessment': [obj, obj, obj, ...]}}

to

{'totaled_scores': {'Assessment1': [obj], 'Assessment2': [obj], ...}

Updated module_tree_reset code to aggregate all of the assessment
objects in this case.

Moved the list of suspended problems to advanced settings (ie
policy.json) instead of defining them in the <randomize> tag within the
course xml. This avoids potentially destroying student state and makes
it easier for course admins to add/remove suspended problems. Course
admins can now define *all* suspended problems in one place simply by
listing them in the 'Advanced Settings' page of the studio web
interface.

LTI grade callbacks for example, come in with an anonymous user.  This causes a
stack trace in the psychometrics app that leads to the LTI service
getting a 404.  This adds a check before that callback gets registered.

Raising an exception in the case that all choices have been exhausted in
the randomize module breaks things so set it to the last seen problem
instead.

Access is now automatically requested on each page load if necessary so
this button is no longer necessary. Updated the template to explain that
access has been requested but not yet granted and to contact the
instructor/TA if they're not given access soon. The page now
automatically checks status every few seconds and will automatically
reload when access has been granted.

The 'position' state is clearly checked by get_student_status, however,
the code never restored 'position' from the state during a reset
before so enabling this now may change behavior. Leaving it
commented for now just in case it's indeed needed later on.

Update get_assignments_attempted_and_failed to fetch only assignments
that have been visited by the student and that were not attempted OR
were failed.

history is the randomize module's complete history. visited will be True
if the user has actually seen the problem (ie the problem was rendered
via the randomize module's student_view). This is indicated by the
choice being in the history given that the randomize xmodule only stores
history entries from student_view.

This way we know the student has seen it given that student_view should
only be called when it's rendered for viewing. Changed the randomize's
history field from a json String to xblock.fields.List as it should have
been in the first place. Randomize history is now always stored
regardless of no_repeats, however, no_repeats must still be specified in
order to prevent duplicates.

Problems can now be suspended by adding a 'suspended' xml attribute to
the <randomize> tag. To suspend a problem simply specify a comma
separated list of problem url_name's in the 'suspended' xml attribute.
Previously setting any attribute at the <problem> level was causing the
system to drop grades for any student that had successfully completed a
suspended problem. This new approach for suspended problem fixes that
issue.

The StudentModuleHistory db entries are filled with incomplete/invalid
entries. Add more checks to ensure the state is complete with attempts,
correct_map, and student_answers and ensure that both grade and
max_grade are not null. There are also some entries whose score doesn't
match the correct_map which we filter by manually counting the
'correctness'='correct' entries in the correct_map and comparing that to
the grade.

Can't compare JSON strings reliably since the order of keys may change
between loads/dumps.

This shows the entire submission history for every problem in a course
for a given user. Takes a course id and student username as input.

Staff users can now navigate to a problem, click 'Proctor Submission
History' button in the proctor admin panel, type a username and view
the user's attempts for each problem they've attempted within the
problem bank.

This is similar to courseware.views.submission_history except that:

1. It works for a list of locations and returns a dictionary rather than a Fragment
2. It filters StudentModuleHistory to those that have 'attempts' in their state.
3. It strips any duplicate (ie identical state) StudentModuleHistory entries.
4. It propagates all exceptions

The proctor server's status call now additionally returns whether the
problem has been requested or not. If it's not been requested yet the
proctor xmodule now automatically submits a request on behalf of the
user. This completely skips the 'request access' page when MIT card
integration is active.

* Use params for GET calls and data for POST

* Update the filter on the list of assignments to reset to include
  assignments that have None for 'earned' or 'attempted'.
* Added --wipe-randomize-history flag to reset_attempts which blasts the
  randomize history for users that have seen/attempted all problems
  (should only be used for testing!)
* Added 'Wipe randomize history?' checkbox to proctor student admin form
* Moved cache/modulestore/signal configuration out of the reset_attempts
  and attempt_stats management command and into module_tree_reset module
* Use student.email instead of student.username + '@mit.edu'

Runs a Python script within the edx platform's lms environment.

* Updated URLs to match latest urls in xproctor repo.
* Use POST for the request call and GET for status.
* Reuse status() in is_released()

(cherry picked from commit 2c8f82f0)

Conflicts:
	AUTHORS

Updated handle_ajax to explicitly handle each proctor server command
individually without accepting any parameters from the client. This
completely avoids users being able to interact with the proctor server
API directly. Updated the JS code to do a POST when requesting access
(TODO: proctor server still needs to be updated to require POST - for
now the proctor xmodule still uses GET under the hood until this
gets changed)

Turns out we can avoid the weird collectstatic issue (extra hash at end
of js filename) by not invoking leanModal from javascript at all.
Instead just use static namespace from 'static_content.html'
and put rel="leanModal" in the <a> DOM element that links to the modal
DOM tree via it's href attribute.

Combined the 'release' and 'disable' dispatch urls in handle_ajax into a
single 'override' dispatch. The new 'override' dispatch expects the user
to POST and toggles the proctor xmodule's staff_release based on the
'enabled' key (true or false) in the request data.