rearrange txeof so it processes slots up to the index given in the

rearrange txeof so it processes slots up to the index given in the
completion event, and only request completion events for the last packet
we place on the ring in bnxt_start, which doesn't do anything now but
will make tx faster when we get tx mitigation working.