LU-17893 tests: wait destroys before replay-dual:test_28

LU-17893 tests: wait destroys before replay-dual:test_28

replay-dual.sh:test_28() should take care that it drops only own
blocking ast. If test_26() ran before, there may be pending destroys
when test_28() runs. Dropping of blocking asts for destroys makes
replay-dual.sh to get accompanied with:

  watchdog stack traces:
  [169376.453554] Lustre: ll_ost00_057: service thread pid 236757 was
  inactive for 40.816 seconds. Watchdog stack traces are limited to 3
  per 300 seconds, skipping this one.
  [169376.461659] [<0>] ldlm_completion_ast+0x99b/0xc00 [ptlrpc]
  [169376.461782] [<0>] ldlm_cli_enqueue_local+0x302/0x890 [ptlrpc]
  [169376.461888] [<0>] ofd_destroy_by_fid+0x29c/0x570 [ofd]
  [169376.461906] [<0>] ofd_destroy_hdl+0x22c/0x960 [ofd]

  lock timeouts:
  [169638.155933] LustreError:
  236757:0:(ldlm_request.c:104:ldlm_expired_completion_wait()) ###
  lock timed out (enqueued at 1729087746, 303s ago); not entering
  recovery in server code, just going back to sleep ns..

  and system overload indications:
  [169852.021044] Lustre: ll_ost00_052: service thread pid 236555
  completed after 516.964s. This likely indicates the system was
  overloaded (too many service threads, or not enough hardware
  resources).

Wait for completion of destroys before starting test_28().

Test-Parameters: trivial testlist=replay-dual
Signed-off-by: Vladimir Saveliev <vladimir.saveliev@hpe.com>
Change-Id: I837579a428d8c2383fe884961d356ff417fc3f2e
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/56712
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Jian Yu <yujian@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>