mac80211: fix powersaving clients races

The code to handle powersaving stations has a race:
when the powersave flag is lifted from a station,
we could transmit a packet that is being processed
for TX at the same time right away, even if there
are other frames queued for it. This would cause
frame reordering. To fix this, lift the flag only
under the appropriate lock that blocks TX.

Additionally, the code to allow drivers to block a
station while frames for it are on the HW queue is
never re-enabled the station, so traffic would get
stuck indefinitely. Fix this by clearing the flag
for this appropriately.

Finally, as an optimisation, don't do anything if
the driver unblocks an already unblocked station.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>