Standard errors in evsi()

[chjackson]

Technically it should be straightforward to add a se argument to evsi(), and handle at least standard errors from the nonparametric regression methods for which standard errors can be calculated in evppi(). Currently this is the GAM, earth and bart methods.

Though the warning should be repeated that these uncertainties are likely to be small in the context of other modelling and VoI uncertainties (e.g. converting individual to population VoI).