On case-insensitive filesystems, check for re-imports more robustly to avoid duplicate database entries

I use beets and iTunes on Windows. iTunes "manages the library" and beets updates tags in files. My library is mostly tagged with Picard, so I've disabled autotag.
The workflow looks like this:
- I've got a file called "04 Pasta de Coca.mp3". The title tag, naturally, reads "Pasta de Coca"
- Now I want to properly tag it and fire up beets. The EP it's on is not in MusicBrainz (actually, it is, but with a somewhat wrong title), so I tag with data from Discogs. It gets the title "Pasta De Coca" - note the case change, - and beets writes this tag to file.
- iTunes notices the change and renames the file into "04 Pasta De Coca.mp3"
- I fiddle with the library a little more; more files get renamed/moved by iTunes
- Then I decide that I need to notify beets of all those moved files and I reimport the entire directory

Now I have only one file, and two tracks with two paths, both of which beets can successfully resolve to that file, so `beet update` doesn't help.

Could have been worse: the first time I reimported everything I typed in the directory name as \music\ instead of \Music\, so I ended up with an entire library duplicated. Had to delete it and restart from scratch.