RDMA/hns: Bugfix for workqueue for flush cqe process

driver inclusion
category: bugfix
bugzilla: NA
CVE: NA

For nvme application scenes, the workqueue should be destroyed after
ib unregister. Otherwise, it will lead to following calltrace:

[  571.164209] Call trace:
[  571.166644]  __queue_work+0x184/0x4e8
[  571.170292]  queue_work_on+0x7c/0x88
[  571.173860]  init_flush_work+0x60/0x70 [hns_roce]
[  571.178553]  hns_roce_v2_post_send+0xa18/0xb78 [hns_roce_hw_v2]
[  571.184464]  __ib_drain_sq+0x134/0x1c8
[  571.188199]  ib_drain_sq+0x24/0x28
[  571.191587]  ib_drain_qp+0x14/0x40
[  571.194976]  nvme_rdma_stop_queue+0x6c/0x78 [nvme_rdma]
[  571.200187]  nvme_rdma_teardown_io_queues.part.10+0x50/0xc0 [nvme_rdma]
[  571.206787]  nvme_rdma_delete_ctrl+0x2c/0x50 [nvme_rdma]
[  571.212085]  nvme_delete_ctrl_work+0x54/0x78
[  571.216341]  process_one_work+0x1e4/0x458
[  571.220337]  worker_thread+0x40/0x450
[  571.223986]  kthread+0x12c/0x130
[  571.227201]  ret_from_fork+0x10/0x18
[  571.230763] Code: aa1c03e0 97ffff8e aa0003e2 b5fff900 (f9400262)
[  571.236844] ---[ end trace ada52cff7b676b94 ]---
[  571.241458] note: kworker/u259:0[4686] exited with preempt_count 1

Signed-off-by: Yixian Liu <liuyixian@huawei.com>
Reviewed-by: huangdaode <huangdaode@hisilicon.com>
Reviewed-by: Xie XiuQi <xiexiuqi@huawei.com>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>