Resend sync request with each retry.

[General-Fault]

Fixes issue #28
Clear the input buffer and resend the sync request with each sync retry for STK500v1.
This partially reverses the change to commit 7765e6d, but it brings it closer inline to what avrdude does.

This is the code from avrdude 6.3 for reference:

int stk500_getsync(PROGRAMMER * pgm)
{
  unsigned char buf[32], resp[32];
  int attempt;

  /*
   * get in sync */
  buf[0] = Cmnd_STK_GET_SYNC;
  buf[1] = Sync_CRC_EOP;
  
  /*
   * First send and drain a few times to get rid of line noise 
   */
   
  stk500_send(pgm, buf, 2);
  stk500_drain(pgm, 0);
  stk500_send(pgm, buf, 2);
  stk500_drain(pgm, 0);

  for (attempt = 0; attempt < MAX_SYNC_ATTEMPTS; attempt++) {
    stk500_send(pgm, buf, 2);
    stk500_recv(pgm, resp, 1);
    if (resp[0] == Resp_STK_INSYNC){
      break;
    }
    avrdude_message(MSG_INFO, "%s: stk500_getsync() attempt %d of %d: not in sync: resp=0x%02x\n",
                    progname, attempt + 1, MAX_SYNC_ATTEMPTS, resp[0]);
  }
  if (attempt == MAX_SYNC_ATTEMPTS) {
    stk500_drain(pgm, 0);
    return -1;
  }

  if (stk500_recv(pgm, resp, 1) < 0)
    return -1;
  if (resp[0] != Resp_STK_OK) {
    avrdude_message(MSG_INFO, "%s: stk500_getsync(): can't communicate with device: "
                    "resp=0x%02x\n",
                    progname, resp[0]);
    return -1;
  }

  return 0;
}

As you can see, it does a send-drain-send-drain then goes into a send-read loop. When I replicated this pattern exactly, the next read would timeout. I'm not sure what that's about - but a drain-send-read loop addresses the problem as far as my testing went.

[christophediericx]

It is an absolute pleasure to receive a pull request which is as clearly described and provides the amount of detail as presented here: awesome work. I'll merge the pull request right now, but I will only have time to fully test it (and prepare a new release) over the course of the next couple of days (in the weekend at the latest).

Thanks!