KH270: avoid wrapping pixel modulo 256

Using modulo 256 arithmetic caused the computed solenoid number to
jump when crossing the left edge of the pattern.
This caused corruption on the leftmost needles of the bed.

src/ayab/knitter.cpp

@@ -435,7 +435,22 @@ bool Knitter::calculatePixelAndSolenoid() {
     return false;
   }
 
-  m_pixelToSet = m_position - startOffset;
+  // Unsigned 8-bit arithmetic computes modulo 256 — this is not
+  // appropriate when you have 12 solenoids, it causes pixel -1 to
+  // have more than 1 solenoid of difference from pixel 0.
+  // So instead we use a 16-bit int to compute the pixel, then
+  // convert it back to an unsigned 8-bit integer using a multiple
+  // of the number of solenoids as the modulus.
+  // We only handle the underflow case, because the machine with 12
+  // solenoids (KH270) has only 112 needles and can therefore never
+  // have positions that cause an 8-bit overflow.
+  int pixelToSet = (int)m_position - startOffset;
+
+  if (pixelToSet < 0) {
+    pixelToSet += Machine_t::Kh270 == m_machineType ? 252 : 256;
+  }
+
+  m_pixelToSet = pixelToSet;
 
   if (!beltShift) {
     m_solenoidToSet = (m_pixelToSet + bulkyOffset) % SOLENOIDS_NUM[static_cast<uint8_t>(m_machineType)];