Better time handling

[nwf]

Right now, we rely on SNTP being fairly prompt and reliable. If it isn't, we'll start off life with a timestamp of zero (and publish that in our heartbeats) and that doesn't seem like it's going to work out well. We should probably subscribe to timesync messages and either check against SNTP or use them as a fallback if we haven't heard otherwise.

[timparenti]

I like the idea of using the timesync messages as a fallback until we've heard from SNTP. Is there a straightforward way for the devices to indicate back to the server which time source they're currently using?

[nwf]

We've never seen SNTP fail to synchronize within a second or two of getting an IP address; the devices hit the CMU NTP server. But, that said, we can certainly add a field to the heartbeat messages to tell us the time source or deviance of local clock at the last timesync message we saw.