Modify samuelson

[shizejin]

Closes #207

This PR corrects several minor errors in the simulation when using the LinearStateSpace class.

1. the attribute of the LinearStateSpace for controlling the initial distribution is mu_0 instead of μ_0
2. to be less confusing, I changed all the attribute names related with LinearStateSpace to their names denoted in the source code
3. when setting the initial distribution to be the stationary distribution, we should use mu_x and Sigma_x instead of mu_y and Sigma_y
4. when the "stationary" mode is off, the code still unintentionally simulates data from its stationary distribution, as we were setting μ_0 at the stationary values. I changed μ_0 slightly so that the simulation shows a convergence to the stationary distribution from somewhere away

[jstac]

Thanks @shizejin ! Does this close #207 ?

[shizejin]

Thanks @shizejin ! Does this close #207 ?

Hi @jstac, yes it does! I have added a linking to the PR to close the issue when merged.

[thomassargent30]

To me this looks very good. Thanks Zejin.