default M10 standby power is high and noisy

[Selene ToyKeeper]

In a fresh-flashed state with wifi enabled, the M10's power usage is higher than it should be, and noisy. The average power draw ranges from 60mA to 90mA, meaning a standby time of 3 to 5 days. However, the momentary power draw rapidly jumps between ~10mA and ~200mA. This is similar to how krillin behaved in early versions.

Turning off wifi reduces standby power to a steady 12mA, for 25 days of standby time.

I tested recent rc-proposed images (190 to 198) and rc images (13 to 18), and the behavior is consistent. There is likely a kernel-level wifi bug similar to what krillin used to have.

Steps to reproduce the issue:
1. Flash a fresh image, wiping everything.
2. Connect to a wifi access point.
3. Let the device go into standby mode.

Expected results: 12mA to 24mA average power (at a rough guess, up to 2X as high as with wifi off).

Actual results: 60mA to 90mA average power.

[Selene ToyKeeper (toykeeper)]

Graph of power use for three test runs, using rc-proposed 198. Shows moment-to-moment power use. Logs are available here:
http://people.canonical.com/~platform-qa/power-results/2016-09-27_21:42:49-frieza-198-power_usage_idle/

[Selene ToyKeeper (toykeeper)]

Average power results for the last few rc-proposed images. Shows variability per image.

Average results for other power tests are here for comparison:
http://people.canonical.com/~platform-qa/power-results/graphs/2016-09-28/frieza.html

[John McAleely (john.mcaleely)]

I appreciate this is reporoduced on non-xenial rc images, but I think it makes sense to look at this while we're doing the xenial bringup, and then port any fixes found.