The logic on the quick open was overcomplex, and it led to
bugs while filling the document list.
The new logic fills a vector with all the files,
organizes them by Url, remove the files based on the
duplicated url's (so name matching doesn't matter) and only
then starts to populate the view.
I belive this is also faster than the old code as I minimized
the amount of memory allocations and deallocations.
I'm unsure about the use of QFileInfo in the end of the code as
we can have too many files there and creating a temporary
QFileInfo just to get the name is a huge overhead, we can do a split('/').last()
for the file name. I'll do some measurements with that.