Easier factory generator function for NumPy binary ufuncs

It is now much easier to roll up your own streaming reduction function using ireduce_ufunc.

Some routines were also added:

• imin, imax
• iinner, idot, itensordot, ieinsum

I'm probably forgetting some more.

Better checks and more streaming functions

In this release, new functions are available: iall, iany, iflatten, and istack. Also, all streaming functions are now safe
to use with streams of non-arrays.

Better axis handling

Reductions along axis is not equivalent to NumPy's routines. It is not possible to reduce along axes that are perpendicular to the 'stream axis', i.e. the stacking direction.

This means is it now possible to replicate the results of numpy functions (sum, prod, average, std, etc...) exactly, no matter the shape of the stream arrays.

NaNs and parallelization

NaNs are now handled in the statistics functions. Parallelization utilities are also available.

Initial release

This initial release contains a few statistical functions and utility functions. This is meant as a starting point.