Improved noise tolerance on recv

[phooky]

Sorry, I made this fix a bit back but forgot to make another PR. This is mostly cleanup and a few improvements on the receiving end's handling of not-so-clean connections. (In particular it handles cases where some noise is sent to the receiver before good data comes through; the hardware I was receiving from was sending some spurious bytes at the start of transmission that lrzsz could handle but xmodem.rs would fail on.)

[awelkie]

So it looks like the main change here is that if the receive function gets an unexpected byte when it expects a packet header, we'll just keep trying instead of considering that an error. I'm not sure if this is what we want to do, though, since it seems like we should treat unexpected bytes as something bad. For example, it looks like if the transmitter sends the same unexpected byte over and over, this receive function will never return.

It seems like the right way to handle this is to allow the user some control over the types of errors that are allowed. So the user could specify to quit after some number of CRC errors, some number of packet timeouts, and maybe an infinite number of unexpected bytes. Or maybe the user could supply a callback between each packet to determine at run-time whether to quit the receive.

Thoughts?