ClientState checkForActivity calculates with nanos and millis

Bug is at least in 1.2.1.

The ClientStates checkForActivity method calculates e.g. the nextPingTime. In one of the else-cases (l. 787) it does:
nextPingTime = Math.max(1, getKeepAlive() - (time - lastOutboundActivity));
getKeepAlive delivers ms, while time and lastOutboundActivity are in ns. Therefore the result of this function is always 1, resulting in high frequented ping schedules.

PS: I strongly recommend naming all time variables with Ms/Nanos in the name to prevent such things.
E.g. in line 718, long nextPingTime = this.keepAlive; is done. this.keepAlive is in nanos, however, nextPingTime in the end has to be in ms again. It is always overwritten in the rest of the method, but this is also confusing.