Allow user to fit glm of their choice in estimateEffects()

[nielsenrich]

We need an option to run effects models besides lm. We used to have this in early code I have before we packaged it up. I'm not sure why it got cut, but probably because the Athey-Imbens intervals don't work or something like that. This is related to the issue of making the Athey-Imbens intervals an option rather than a requirement.