Create timeseries re-binning common primitive

[cdbethune]

Add a primtive to common primitives to allow for timeseries data to be re-binned using different time buckets. At minimum, we should support seconds, minutes, hours, days, weeks, months, years.

[cdbethune]

Timeseries data will be passed into the binning primitive in the agreed upon D3M long format, where data is stored in a single dataframe, with each series having a unique identifier associated with it. The series values themselves will be a combination of a date/time or sequence value, and a set of measures.

The primitive should use the following parameters:

Column Parameters:

• key column - when unset should use the first column found that has the GroupingKey semantic type as the ID column
• timestamp column - when unset use the first column with the DateTime semantic type as the timestamp column
• value columns - any columns with Integer or Real types as value columns Columns that are not part of the above set should be dropped.

Binning Parameters:

• granularity - selected from an enumeration of seconds, minutes, hours, days, weeks, months, years
• binning operation - selected from an enumeration of sum, mean, min, max

The binning process will bin the data for a given series ID to the granularity requested using the supplied operation. All bins must be returned - if no data is present for a given bin it should be set to NaN.