Don't allow submitting of huge files

@ellisonbg apparently had a student submit a 1TB file which filled up the drive of the server and crashed it. It seems like it would be worthwhile for nbgrader submit to check the file size and either exclude files that are very large or at least prompt the user “are you sure?”

[rkdarst]

Deduplicating sounds interesting but as jhamrick says, easier said than done.

This made me realize there are two use cases (at least for us), which means there may need to be two separate ways to configure:

Copying data files to the student, which are needed to run the assignment. My recommendation to teachers for our system is to not do this, use a shared course data directory that is mounted in all student environments. However, this makes copying assignments out for running harder because the assignment directory is not self-contained.

Second is coping things back to the instructors. We have one course whose notebooks train machine learning models and write the models to files. These need to be copied back to the instructors (but not the data files!).

And of these cases, none of them need to be copied back yet again for the feedback step.

So for each of these steps, there are possible size limits, blacklists, whitelists, etc…

(And as a bonus, make the student-side and possibly insturctor side configurable without nbgrader_config.py… this configuration somehow embedded in the assignment directories…)

[rkdarst]

Note from @sefriol: if data gets copied to students and back, students can modify it which may affect the grading process…