LU-11912 tests: SEQ rollover fixes

LU-11912 tests: SEQ rollover fixes

To avoid changeing SEQ after replay_barrier, we
use force_new_seq when starting the test suites heavily
using replay_barrier, e.g. replay-single.
However when there are fewer OSTs, the default 16384
SEQ width could not last the entire test suite, SEQ
rollover could still happen randomly after replay_barrier.

To overcome this, change the default OSTSEQWIDTH to
65536, and divide by number of OSTs, so the SEQ width is
larger with fewer OSTs. For 8 OSTs, the SEQ width is 16384,
and make sure we don't go under it.

Use force_new_seq_all for the test suites using replay_barrier
on MDTs other than mds1.

Add force_new_seq_all for replay-ost-single, which is using
replay_barrier on OST. If SEQ rollover happens after that,
the SEQ range update on ofd is lost due to replay_barrier,
the next time when we try to allocate a new SEQ will end up
with an old one.

Use force_new_seq_all for the test cases(namely sanity-pfl/0b
0c 1c 16b sanity/27Cd) checking for number of stripes created
with overstriping, to make sure we have enough objects
in the precreate pool.

Test-Parameters: trivial ostcount=4 testlist=replay-single
Test-Parameters: ostcount=2 testlist=replay-single
Test-Parameters: mdtcount=2 testlist=conf-sanity env=ONLY=122a,ONLY_REPEAT=10
Test-Parameters: testlist=sanity,sanity-pfl
Test-Parameters: testlist=sanity-scrub,replay-single,obdfilter-survey,replay-ost-single,large-scale
Fixes: 0ecb2a167c ("LU-11912 ofd: reduce LUSTRE_DATA_SEQ_MAX_WIDTH")
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Change-Id: Ic65111199f042405d6db8acb729b2cddf91138af
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/50478
Tested-by: Maloo <maloo@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>