Changes:
- use high resolution timer
  latency became approx. 10-15 times less than before
- use taklet for sending remaining data in tx path
Tested on Intel/ARM.
Reviewed-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Tested-by: NDmitry Tarnyagin <Dmitry.Tarnyagin@stericsson.com>
Signed-off-by: NAlexey Orishko <alexey.orishko@stericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>