Another failure of decoding UAT

[Pete Zaitcev]

Guys,

As you know, our sample rate is not good enough to decode UAT's FSK in
a conventional way, so I thought to try something else: use 2 samples
per bit to detect the frequency shift. This is obviously unreliable,
but I hoped to see something at least. But it didn't work: seems like
decoding noise.

The code is here:
 https://github.com/zaitcev/ruat

Anyone is willing to critique? Where did I go wrong?

Thanks,
-- Pete