LU-5242 osd-zfs: umount hang in sanity 133g

LU-5242 osd-zfs: umount hang in sanity 133g

It's not optimal to free large/sparse object directly by
dmu_object_free(). The dmu_tx_count_free() may over estimate the
memory overhead and cause the TX to fail in arc_tempreserve_space()
and eventually hang.

The correct way is to truncate the object with dmu_free_long_range()
before freeing it. This patch moves large object to ZPL unlinked set
and then later truncates and frees it, with as many TXs as needed.

Change-Id: I077dbcafff5abaea4a42dc9bc3f9ae980bdcdc54
Signed-off-by: Isaac Huang <he.huang@intel.com>
Test-Parameters: envdefinitions=SLOW=yes mdtfilesystemtype=zfs mdsfilesystemtype=zfs ostfilesystemtype=zfs
Reviewed-on: http://review.whamcloud.com/13630
Tested-by: Jenkins
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>