Allow predict.glmpca to infer factors from new data given a trained model

[jakeyeung]

It would be great to use a trained glmpca to infer factors from new data.

Possibly like:

new_factors <- predict(glmobject, data = new_data)

I imagine the inference part would involve fixing the loadings to be constant, while inferring the factors from the new samples, but there might be details I'm missing.