Previously, api.update_from_assessments() would call model.
update_from_assessments(), then api._serialized_with_details(reqs), which
would duplicate the update work we *just* did as part of getting the
status. Now, update() and get_details() are totally seperate operations, and
_serialized_with_details() is a read-only method that does not need to have
requirements passed in.

I've also standardized student_training and peer, the two assessment types that
do *not* auto-complete on None requirements, to return False if None, and error
out if passed {}. This is to allow staff_assess to pass None to indicate
"Update thyself, if you don't depend on requirements" to the other step types.

Also adding test_requirements_changed(), to confirm that workflows are updated
when requirements change, even though thoe requirements are no longer explicitly
passed to the getter.

Fairly simple test, should help to provide some extra peace of mind.

I misunderstood what was happening in my cursory investigation earlier -
all that needs to change here is the type of exception being caught.

TNL-3464

Add a test to verify that it is working as expected.

Staff assessments can now mark workflow as being done, with proper handling
for blocking workflows (such as peer), to hide a score if the submitter
has not yet fulfilled their obligations. Includes tests.

Also contains a tiny bugfix for a previously unexecuted codepath.

This test is trying to confirm behavior when an api cannot be loaded.
It is no longer possible for an AssessmentWorkflow to have been created
without going through __init__(), so the latter part of this test is
now trying to confirm assumptions about an impossible state.

I recommend we just remove it.

TNL-3467

Even when the icon is updated to fa, nothing happens when
the icon is clicked.

TNL-3799

Staff assessments can now mark workflow as being done, with proper handling
for blocking workflows (such as peer), to hide a score if the submitter
has not yet fulfilled their obligations. Includes tests.

Previously, I moved some assumptions out of class contstants and into
code, in `start_workflow()`. Turns out, it's also possible to load a model
from the database, which bypasses this check and fails to fill out self.STATUS
properly. Refactoring so that `include_staff_in_class()` gets called in both
cases.

Changes include:
    -modifies requirements to gt an updated version of edx-submissions
    -adds set_staff_score to the AssessmentWorkflow model, which will record
        a new annotated score using edx-submissions functionality
    -prevents recording non-staff scores if a staff score exists
    -modifies update_from_assessment to call set_staff_score as needed
        -this includes changes to both the workflow model and its api
    -modifies get_score to allow staff overrides as optional
    -modifies the assessment serializer to include id information
        -adds this information to get_score method in self, ai, peer, and staff
            apis, to expose c ontributing_assessments where needed
    -fixes a small bug regarding None vs {} in the peer api
    -adds staff-assessment to the xblock, and makes it always available
    -uses the new force_update_score parameter on the workflow api when
        recording a staff assessment

TNL-1162

Increase the comment textarea size to 1000

Fix styling issue with ORA staff tools

TNL-3845

Speed up the translation process

Bump the version number to 0.2.3

Update translations from Transifex

Miscellaneous small improvements

* Bump version numbers
* Introduce JSCS style validation
* Address style failures

fix travis configuration