The move detection introduced an off-by-one where the last line in a

chunk wasn't being processed as a move unless the line after the chunk

had content. That last line should never have been processed either

in the first place, but we were just looping too hard.
We now loop only through the lines in the chunk, and if we've hit the

line and have any move ranges calculated, we process them.

Saw this initially with a diff containing a single line I expected would

have been marked as a move. After this change, it was properly marked.
Unit tests pass. The new test failed without the fix.