Detect and handle hardlinks

[ssokolow]

1. It's a waste of time to read the same inode once for each hardlink associated with it and unnecessarily slows things down if the user re-runs fastdupes later.
2. It doesn't make much sense to list things as duplicates when, presumably, the hardlinking was the chosen method of deduplication.
3. At the same time, it's not immediately obvious how to represent multiple names for the same inode in the output. (Possibly indenting all but the first one?)

I'll need to come up with some proposals than implement a groupBy classifier that operates on inodes. Ideally, sharing cached stat() results with the size classifier.

(Though its output won't behave quite the same way as the others since, instead of determining what should be compared to what in the final full-content examination, it'll be used to determine which paths don't need to be compared at all.)