Not handling millis() rollovers

[jgstroud]

The code is so stable now that my millis() counter actually rolled rover.
https://docs.arduino.cc/language-reference/en/functions/time/millis/

My uptime had mysteriously gone back to 0, which originally I thought there must have been a power flicker. But there was nothing in the logs. Then today I clicked the wall button and the device went into recovery mode after a single press. I went back and reviewed my syslog logs and see that the counter reset timeout value was set before the rollover.

[jameyers4]

I, for one, would appreciate this change incorporated into 1.8.3 at earliest reasonable convenience. I was closing in on 40 days when I had to power down for non RATDGO reasons - I was close (-ish) to the 32 bit counter limit…

[jgstroud]

PR is available. There is no way to artificially test this as far as I can tell. I'm debating whether I should just merge it now or wait 50 days to verify it properly handles rollovers.