Remove magic values for LowPower.sleep()

[polygamma]

As of now, the SigFox library uses kind of magic values when it comes to LowPower.sleep usage, as can be seen here and here. After reading the datasheet from Atmel, here, point 2.2.3 seems to give a hint, where they are coming from, it's still bad to rely on hardcoded sleep times.

This PR removes time based sleep entirely, and waits for the interrupt coming from the SIGFOX_EVENT_PIN only. After some testing, this definitely lead to more accurate return values of SigFox.endPacket.

The slightly unelegant way of using the wakeUpRoutine function is needed, because in deep sleep, interrupts based on rising or falling edges are not possible. The relevant clock is turned off, which means one has to use "HIGH" or "LOW" as trigger, and thus it is needed to change the relevant values in a callback function to avoid an infinite loop. It is further needed to call a status() function without delay, because that leads to a deadlock, when being called from within an interrupt.

Anyway, this does only work with a fixed version of the Low Power library, see arduino-libraries/ArduinoLowPower#9.
As of now, it is not possible to register interrupts for the SigFox event pin.

Btw this commit fixed the problems with the controllers being unable to wake up from LowPower.sleep at all. Using RTCZero including this commit, removes the need to always run with the debugging mode turned on.

[facchinm]

LGTM! I'll test it later today with arduino-libraries/ArduinoLowPower#9, although I'd wait to merge when both the PR are fine.

[polygamma]

The relevant commit to make this work, can now be found here

[polygamma]

Addition, since we've been speaking about out of bounds errors here, you may want to include this commit.

Edit - in total are three commits needed: commit1, commit2, commit3

[polygamma]

Any update?