xfs_repair: correctly detect partially written extents

xfs_repair: correctly detect partially written extents

Recently, I was able to create a realtime file with a 16b extent size
and the following data fork mapping:

data offset 0 startblock 144 (0/144) count 3 flag 0
data offset 3 startblock 147 (0/147) count 3 flag 1
data offset 6 startblock 150 (0/150) count 10 flag 0

Notice how we have a written extent, then an unwritten extent, and then
another written extent.  The current code in process_rt_rec trips over
that third extent, because repair only knows not to complain about inuse
extents if the mapping was unwritten.

This loop logic is confusing, because it tries to do too many things.
Move the phase3 and phase4 code to separate helper functions, then
isolate the code that handles a mapping that starts in the middle of an
rt extent so that it's clearer what's going on.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>