LU-9859 lnet: convert selftest to use workqueues

LU-9859 lnet: convert selftest to use workqueues

Instead of the cfs workitem library, use workqueues.

As lnet wants to provide a cpu mask of allowed cpus, it
needs to be a WQ_UNBOUND work queue so that tasks can
run on cpus other than where they were submitted.
We use alloc_ordered_workqueue for lst_sched_serial (now called
lst_serial_wq) - "ordered" means the same as "serial" did.
We use cfs_cpt_bind_queue() for the other workqueues which sets up the
CPU mask as required.

An important difference with workqueues is that there is no equivalent
to cfs_wi_exit() which can be called in the action function and which
will ensure the function is not called again - and that the item is no
longer queued.

To provide similar semantics we treat swi_state == SWI_STATE_DONE as
meaning that the wi is complete and any further calls must be no-op.
We also call cancel_work_sync() (via swi_cancel_workitem()) before
freeing or reusing memory that held a work-item.

To ensure the same exclusion that cfs_wi_exit() provided the state is
set and tested under a lock - either crpc_lock, scd_lock, or tsi_lock
depending on which structure the wi is embedded in.

Another minor difference is that with workqueues the action function
returns void, not an int.

Also change SWI_STATE_* from #define to an enum.  The only place these
values are ever stored is in one field in a struct.

Linux-commit: 6106c0f82481e686b337ee0c403821fb5c3c17ef
Linux-commit: 3fc0b7d3e0a4d37e4c60c2232df4500187a07232
Linux-commit: 7d70718de014ada7280bb011db8655e18ed935b1

Lustre-change: https://review.whamcloud.com/36991
Lustre-commit: 51dd6269c91dab7543cd9dfd1848c983efa6db36

Test-Parameters: trivial testlist=lnet-selftest
Change-Id: I5ccf1399ebbfdd4cab3696749bd1ec666147b757
Signed-off-by: Mr NeilBrown <neilb@suse.de>
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/56774
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Yang Sheng <ys@whamcloud.com>
Reviewed-by: James Simmons <jsimmons@infradead.org>
Reviewed-by: Oleg Drokin <green@whamcloud.com>