Grade passback logging

[arthurian]

This PR updates the info logging for grade passback so that if an error occurs, we can more quickly locate the course/user that is experiencing the issue and start debugging.

Changes:

• Updated the lti_grade_passback() method so that all info log statements follow the general format of LTI grade request ... lti_log_data=DICT.
• There are 3 classes that implement that method: StoreBackend, CatchStoreBackend, WebAnnotationStoreBackend. It's unclear why the method is duplicated in all 3 classes, but that should be revisited (out of scope to refactor in this PR).

Launch parameters included for debugging:

• context_id: maps to the LTICourse.course_id Django model, so we can use this to identify the specific course that is using the tool.
• user_id: maps to the LTIProfile.anon_id Django model, so we can use this to identify the specific user initiating the request, or at the very least tell if it's the same user or a different user.
• lis_outcome_service_url and lis_result_sourcedid: these are specific to the outcome request and show where the tool is POSTing the grade to. Not sure how useful they will be.
• launch_presentation_return_url: this is not strictly necessary, but may be useful to identify the originating canvas course without using a canvas-specific launch parameter (since this tool is used in edX too). Per the LTI v1.1 spec this is the Fully qualified URL where the TP can redirect the user back to the TC interface. In Canvas, this will look something like this: https://canvas.harvard.edu/courses/:COURSE_ID/external_content/success/external_tool_redirect.

Outcome response data:

• In addition to the launch parameters added to the log for context, we can also get at the underlying POST response that is sent to Canvas. This is tucked away in the OutcomeResponse object which is returned from the tool_provider.post_replace_result(score) call.
• See the post_outcome_request() method in lti.outcome_request.OutcomeRequest.

[arthurian]

After digging a bit, looks like we can get the raw response code and content (if any), which might be useful since the outcome.description is being logged as None, which means the XML parsing is throwing an exception somewhere.