MEKF redesign

[vegard-solum-4ss]

Description

Refactor MEKF implementation for readability and speed (AHRS/VRU mode).

Runs >70% faster than current implementation.

Key improvements:

• Calculate state transition matrix and process noise covariance directly, avoiding the intermediate continuous-time state space matrices.
• Remove the need for prior state and covariance
• Simplified Kalman filter state and covariance update functions.
• Take coning and sculling integrals as input instead of instantaneous rate measurements. Note that rate measurements can still be applied if desired using dvel = f * dt and dtheta = w * dt.
• Speed-up attitude reset by exploiting the renormalization.

Checklist

• PR title is descriptive and fit for injection into release notes (see tips below).
• Correct label(s) are used.

PR title tips:

• Use imperative mood.
• Describe the motivation for change, issue that has been solved or what has been improved - not how.
• Examples:
  • Add functionality for Allan variance to smsfusion.simulate
  • Upgrade to support Python 3.10
  • Remove MacOS from CI