Ensure Euler-Maruyama compatibility with Gaussian-transition filters

[mattlevine22]

1. Check compatibilities
2. Does the EM-discrete-evolution need to be a special class to be recognized for structured inference? If so, fix this.