Biggus

Highlights:
- 184 Python 3 source compatibility
- 193 test with numpy 1.9
- 176 Failing type return on masked array on LHS
- 195 197 Replace vectorize with list comprehension in ArrayStack.**deepcopy**

Highlights:
- 151 NumPy 1.10+1.11 fixes
- 173 Np slice stack deepcopy
- 177 Namespace tidy
- 183 Re-add ArrayContainer to public API
- 181 Add functionality to create a lazy masked array where equal to a given value

Highlights:
- Add support for all NumPy ufuncs and a few more besides (153).
- Add __r*__ operators (155).
- Fix indexing of broadcast dimension (156).
- Fix ArrayStack for Fortran-order arrays (159).
- Fix indexing of unary operations (162).
- Fix the handling of masked arrays for unary operations (167).
- Fix the ability to use binary operators on two different bit-length data types (165).
- Mimic NumPy's special type-promotion behaviour when dealing with Python floats & ints (169).

Highlights:
- A fix for the `__getitem__` for `LinearMosaic` that corrects array slicing output (147).
- Biggus' `__div__` method properly accounts for the dtype of arguments passed to it (149).
- Added support for Python 3.5 (150).

Highlights:
- Fix to ndarray indexing on LinearMosaic (144)
- Support for dtype promotion on elementwise operations (145)
- Addition of `__orthogonal_indexing__` protocol (142)
- Addition of the astype array method and associated AsDataTypeArray object (141)
- Improved str and repr representations of arrays (139)

Highlights:
- `biggus.multiply()`, `biggus.floor_divide()`, `biggus.true_divide()`, and `biggus.power()`
- Operator support for `+`, `-`, `*`, `/`, `//`, and `**`.
- Improved broadcasting.