Allow forecasting with exogenous variables (pymc-devs#372)

* Allow forcasting with exogenous variables

* Improve scenario forecasting test

* Wrap user scenarios in dictionary

* Better handling of start date

* Add `append_x0` argument to `LinearGuassianStateSpace`

* Respect `periods` exactly when provided

* Improve forecast-related tests

* Expand test coverage of forecast option combinations

* Make IRF robust to degenerate covariance matrices

Scale rotated covariance matrices to have unit std

* VARMAX parameters had wrong shape

* Solve instead of inverse in kalman step

* Check correct regex in validation test

* Another regex fix

* Remove shutil warning filter

Author: jessegrabowski