The first point is that you have to accumulate (integrate) in small enough units that you don't lose precision. Then when you have accumulated enough, you can convert to degrees.
Each unit of sensor input is worth a fraction
of a degree, (~1/53).
Code:
// 27 mV * sec 1V 127 1 (loop)
// ----------- * -- * ---- * ----
// degree 1000 mV 2.5V 0.026 sec
So you can accumulate the numbers from the sensor (with 127 meaning zero), then if you have
more than +53, thats 1 degree....
Also multiply by "delta_t + 1" to account for possible added delay.
You could use many other forms of multi-byte math, they are all really doing fixed point math with some scale factor using integers.