Add summary statistic functions

[ha0ye]

For the purposes of computing MASE (mean absolute scaled error), we need mean and sd for the time series (and not just the observations that are in the outputs of the forecasting functions).

This is best done as some sort of summary function to be applied to each time series (I'm currently waiting to see what the summary function in weecology/MATSS ends up looking like), and then wrapped in code to process the raw outputs to generate forecast metrics.

So two tasks here:

• [x] summary function for individual time series
• [ ] function to compute forecast metrics given forecast outputs and summary function outputs

[ha0ye]

With the change in #16, the error component for MASE calculations is embedded into the hindcast() function, so this is mostly resolved. Other forecast metrics are better handled in the roadmap for defining forecast evaluation metrics.