Fix filenames in GridFS

[flavioamieiro]

Some titles are unique in our original dataset but are not when we save them in GridFS. This happens because what is valid as a filename in GridFS is a subset of what is valid as the title of an article in Wikipedia. There is an issue to fix this for new uploads (NAMD/pypln.web#89) but we should create a script to fix current filenames and avoid uploading everything again.

[turicas]

I think with this issue we should check and delete duplicated filenames, since the script that will fix the problem will iterate over the set of duplicated filenames.