Fixed sensor.accelerometer gravity sign for consider_gravity = True case

[vapsik]

Issue

In the case of consider_gravity = True the current sensor.accelerometer sensor accounts for gravity vector incorrectly — $$a_{ \text{inertial} } = a_{\text{ENU}} + g$$ but since the gravity vector has components $g = (0,0,-9.8)$ in ENU frame then it must be $$a_{ \text{inertial} } = a_{\text{ENU}} - g$$.

A simple example showcasing the current implementation being incorrect is with a stationary state (1) and airless free-falling rocket (2) - both with ENU-aligned accelerometers: 1) during stationary phase where $a_{\text{ENU}} = 0$ the consider_gravity = True accelerometer reading shows $a_{ \text{inertial} } = (0,0,-9.8)$ where as real IMU would show $a_{ \text{inertial} } = (0,0,+9.8)$ as normal force upward from the ground must negate the effect of gravity for achieving the stationary state. 2) as a rocket goes into (airless) free fall real IMU would show inertial acceleration of 0 (weightless) but with current implementation in free fall or near apogee before parachute opening the consider_gravity = True accelerometer shows total acceleration of nearly 19.6 showcasing the wrong gravity sign usage.

Changes

In accelerometer.py changed inertial_acceleration = Vector(u_dot[3:6]) + gravity into inertial_acceleration = Vector(u_dot[3:6]) - gravity.

Test file showcasing the bugs (made with Claude)

repro_gravity_bug.py