wifi: mt76: use atomic iface iteration for pre-TBTT work

In addition to the previous series I posted, over time I'd
also like to get rid of the iflist_mtx in mac80211. That
isn't easy now since lots of places use iteration and would
have to be audited, but even a cursory look suggests that
mt76 might be more problematic than most since holding the
wiphy lock for the latency-sensitive pre-TBTT work could be
an issue.

Convert the pre-TBTT work to use atomic iteration and then
sending the device commands outside of it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>