Sometimes unable to compute CPU percentage at measurement node startup

[mm318]

This has been brought up at #40 (comment) and #59 (comment).

Currently, the first measurement occurs at the end of the first measurement period (which is also the start of the second measurement period).

If the measurement period and publishing period are at about the same magnitude, there may have been only one measurement at the time of calculating the first statistic message for publishing. This is a problem for the CPU usage measurement nodes, because calculating CPU usage requires having two measurements.

The solution would be to make the first measurement at the start of the first measurement period. (For consistency, we might as well do this for the memory measurement nodes as well.)

[mm318]

If the user for some reason sets the measurement period to be greater than the publishing period, then all bets are off.