Wrap bottleneck for fast moving window aggregations

[shoyer]

Like pandas, we should wrap bottleneck to create fast moving window operations and missing value operation that can be applied to xray data arrays.

As xray is designed to make it straightforward to work with high dimensional arrays, it would be particularly convenient if bottleneck had fast functions for N > 3 dimensions (see pydata/bottleneck/issues/84) but we should wrap bottleneck regardless for functions like rolling_mean, rolling_sum, rolling_min, etc.

[shoyer]

Wrapping bottleneck for operations like nanmean was fixed in #304.

[shoyer]

Some further thoughts:

• I would probably implement these as methods on Dataset/DataArray (e.g., ds.rolling_mean(dim='time', window=5) instead of functions. This is different from pandas, but seems slightly more in line with how everything else in pandas/xray operates.
• I don't have a strong opinion whether we use the name rolling (like pandas) or moving/move (like bottleneck). Maybe rolling is slightly more obvious?
• These functions from bottleneck could be easily extended to work on dask arrays by using using ghost: http://dask.pydata.org/en/latest/ghost.html

[shoyer]

There's been some related discussion about a better API for the rolling aggregation functions in pandas recently: pandas-dev/pandas#10702

[jhamman]

closed via #668