Add optional Kalman update diagnostics

[FlorianPfaff]

Summary

• add normalized innovation squared diagnostics to the backend-native linear-Gaussian update primitive
• allow optional diagnostics return with residual, NIS, measurement-noise scale, and action label
• forward diagnostics through KalmanFilter.update_linear, update_identity, and update_model without changing default return behavior
• add Kalman filter regression tests for diagnostics and scaled measurement noise

Testing

• python -m py_compile src/pyrecest/filters/_linear_gaussian.py src/pyrecest/filters/kalman_filter.py tests/filters/test_kalman_filter.py locally on the generated files
• full pytest not run locally because the container cannot clone/install GitHub dependencies