btrfs: read key again after incrementing slot in move_existing_remaps()

btrfs: read key again after incrementing slot in move_existing_remaps()

Fix move_existing_remaps() so that if we increment the slot because the
key we encounter isn't a REMAP_BACKREF, we don't reuse the objectid and
offset of the old item.

Link: https://lore.kernel.org/linux-btrfs/20260125123908.2096548-1-clm@meta.com/
Reported-by: Chris Mason <clm@fb.com>
Fixes: bbea42dfb91f ("btrfs: move existing remaps before relocating block group")
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Mark Harmstone <mark@harmstone.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>