We should support more aggregations of codes and numerical values

[mmcdermott]

Additional aggregations of possible interest include:

• value/last (the value at the end of the window being summarized)
• value/first (the value at the start of the window being summarized)
• code/time_since_last (how long has it been since the last occurrence of the code)
• code/time_since_first (how long has it been since the first occurrence of the code)
• code/min_time_between (what is the minimum time between occurrences of this code in the window)
• code/max_time_between (what is the maximum time between occurrences of this code in the window)
• value/temporal_slope (what is the slope of the line of best fit between the values observed and the timepoints at which they were measured)?
• value/residuals_vs_linear/mean (what is the mean size of the residual vs. the temporal line of best fit of the values vs. time)?
• value/trendline_crossings/count (how many times does the value graph "cross" the trendline w.r.t. time)

Probably many others to consider as well.