Compliance test needs clock tolerance of 5%

[terrillmoore]

In working on #215 and preparing to merge, I found that a clock tolerance of zero is just too tight; empirically I found that a tolerance of 5% was adequate. Added to the test script by 1b463a3, but I wonder if we should really default to 5% in the LMIC code.

[terrillmoore]

In chasing the remaining FSK issues, I tried turning off clock tolerance in the compliance script, and things worked much better for RX1. RX2, it’s not clear. But rather than starting before the RX2 window, the lmic starts a little later, at about 2.01 sec. RX1 starts consistently at 999. Curious.