Increase precision for percent changes

Most notably with small percent changes, add or subtract, the error
introduced by truncation on double to integer keeps adding up.

Round raw values calculated from percentages to reduce error introduced
by truncation from double to integer.

src/light.c

 
 static bool _light_percent_to_raw(light_device_target_t *target, double inpercent, uint64_t *outraw)
 {
-    uint64_t max_value = 0;
+    uint64_t max_value = 0, target_value;
+
     if(!target->get_max_value(target, &max_value))
     {
         LIGHT_ERR("couldn't read from target");
         return false;
     }
 
-    double max_value_d = (double)max_value;
-    double target_value_d = max_value_d * (light_percent_clamp(inpercent) / 100.0);
-    uint64_t target_value = LIGHT_CLAMP((uint64_t)target_value_d, 0, max_value);
-    *outraw = target_value;
-    
+    inpercent = light_percent_clamp(inpercent);
+
+    /* Round to the nearest integer by adding 0.5 after division.
+       Reduces error introduced by truncation after multiple increments or
+       decrements. */
+    target_value = max_value * inpercent / 100.0 + 0.5;
+    *outraw = LIGHT_CLAMP(target_value, 0, max_value);
+
     return true;
 }