Improve sync word detection

[mweyland]

Detecting the sync word is currently solved trivially. There is two ways we could improve on that:

1. Compute the hamming distance between the template and the masked shift register and allow for a slight mismatch. The threshold parameter has already been prepared in case we need this.
2. Only match e.g. the last 3 bytes of the sync word instead of the full 5 bytes. This is useful if clock recovery fails while transitioning from the preamble to the sync word. This seems to happen if the loop gain of the block is slightly off...