AP_Compass: specify floating point constants

Probably indicates compiler floating-point-constants directive isn't
working

libraries/AP_Compass/AP_Compass_MMC3416.cpp

@@ -220,7 +220,7 @@ void AP_Compass_MMC3416::timer()
             have_initial_offset = true;
         } else {
             // low pass changes to the offset
-            offset = offset * 0.95 + new_offset * 0.05;
+            offset = offset * 0.95f + new_offset * 0.05f;
         }
 
 #if 0

libraries/AP_Compass/Compass_PerMotor.cpp

@@ -117,7 +117,7 @@ float Compass_PerMotor::scaled_output(uint8_t motor)
     uint16_t pwm = hal.rcout->read_last_sent(motor_map[motor]);
 
     // get 0 to 1 motor demand
-    float output = (hal.rcout->scale_esc_to_unity(pwm)+1) * 0.5;
+    float output = (hal.rcout->scale_esc_to_unity(pwm)+1) * 0.5f;
 
     if (output <= 0) {
         return 0;

libraries/AP_Compass/Compass_PerMotor.h

@@ -21,7 +21,7 @@ public:
     
     void set_voltage(float _voltage) {
         // simple low-pass on voltage
-        voltage = 0.9 * voltage + 0.1 * _voltage;
+        voltage = 0.9f * voltage + 0.1f * _voltage;
     }
 
     void calibration_start(void);

libraries/AP_Compass/Compass_learn.cpp

@@ -44,7 +44,7 @@ void CompassLearn::update(void)
 
         // setup the expected earth field at this location
         float declination_deg=0, inclination_deg=0, intensity_gauss=0;
-        AP_Declination::get_mag_field_ef(loc.lat*1.0e-7, loc.lng*1.0e-7, intensity_gauss, declination_deg, inclination_deg);
+        AP_Declination::get_mag_field_ef(loc.lat*1.0e-7f, loc.lng*1.0e-7f, intensity_gauss, declination_deg, inclination_deg);
 
         // create earth field
         mag_ef = Vector3f(intensity_gauss*1000, 0.0, 0.0);
@@ -213,7 +213,7 @@ void CompassLearn::process_sample(const struct sample &s)
             predicted_offsets[i] = offsets;
         } else {
             // lowpass the predicted offsets and the error
-            const float learn_rate = 0.92;
+            const float learn_rate = 0.92f;
             predicted_offsets[i] = predicted_offsets[i] * learn_rate + offsets * (1-learn_rate);
             errors[i] = errors[i] * learn_rate + delta * (1-learn_rate);
         }