mlx4_en: Removed TX locking when polling TX cq

There is no need to synchronize the polling with the transmit
function. The only place to synchronize is when we process
the cq from the transmit function. Also removed spin_lock_irq,
and using spin_trylock, if somebody else is already processing the cq,
no need to wait for it to finish.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>