Missing data handling

This is a re-submission and a maintenance release. The main changes are a small tweak to how missing observations are handled in the function preprocess_SCD() and a corresponding modification to one of the example datasets.

• Modified handling of missing outcome observations in preprocess_SCD() so that the session-by-treatment interaction variable is computed prior to dropping missing outcome observations.
• Modified Bryant2018 example dataset to clarify the timing of treatment phases.
• Added a newdata argument to graph_SCD(), which allows for computing fitted values from model_fit based on a different design structure than the actual data.
• Updates to package documentation.