Add filtering and more prediction types to the SLAM tracker

This MR improves the SLAM tracker with various prediction and filtering methods that can be configured at runtime. More experimentation in different environments, light conditions and types of tracked devices is needed to pick the best methods and their parameters.

• Predictions: Added two kind of predictions, one that integrates gyro samples for angular velocity, and other that does the same with accel samples and linear velocity.
• Filtering: Added three kinds of trajectory filters; moving average, exponential smoothing and one-euro.