LU-18392 tests: hold group locks in recovery-small/160

LU-18392 tests: hold group locks in recovery-small/160

recovery-small/160 starts 10 threads and holding group locks
for 20s.
The osp destroy retries could actually happen on ofd after the 20s
delay, and make those rpc waiting for OST commit while increase
destroys_in_flight.

Make sure we hold on to the group locks while checking for
destroys_in_flight, and add additional check to make sure
the object destroy actually done after releasing the group lock.

Do not cast the transno to 32bit unsigned in osp_sync_interpret(),
which could produce a confusing transno in debug log.

Test-Parameters: trivial testlist=recovery-small env=ONLY=160,ONLY_REPEAT=100
Fixes: 27f787daa7 ("LU-15737 ofd: don't block destroys")
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Change-Id: I2a6ac9700a3e79e9930cee905c1da73da948ba1a
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/56846
Reviewed-by: Alexander Boyko <alexander.boyko@hpe.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>