Viscid

======

Changes:
- xl and xh properties added to Fields and AMRFields
- Fields now have get_slice_extent

Bugfixes:
- XDMF reading uniform crds was wrong; ggcm is using zyx order for O+dxdydz since that seems to work for Paraview, but the xdmf docs call for xyz order, so something isn't kosher here but I guess this works for now
- xyz ordering for scalar_fields_to_vector
- TypeError on numpy reduction operations on ScalarFields
- fix Volume genr_points for xyz ordering

======

Backward Incompatible Changes:
- ALL indexing is now in the natural order (xyz as opposed to zyx). This means seeds, lines, fields... everything. Since this is such a big change that unavoidably breaks many things, this is the only feature of this release, so if you have code that depends on zyx ordering, stay at version 0.80.8.

======

Bugfixes:
- vector -> scalar numpy reduction operations

======

Changes:
- Files can now be unloaded (treating a file like a context manager will unload it on __exit__)
- jrrle readers can handle files with ascii in them
- hdf5 datawrappers can specify a component dimension / index
- preliminary gkeyll reader
- add `mpl.auto_adjust_subplots()` which is like tighen, but doesn't change axes limits on you.
- All fields (regular + amr) now have an 'interpolated_slice' method
- Cython Interpolate and Streamline now work on AMR Fields

Bugfixes:
- calling ufuncs with AMR fields
- numpy broadcasting with Fields of different type / shape
- slicing by UT time was relative to 0 instead of the start of the simulation

Backward Incompatible Changes:
- Slices by value are now a string of '[0-9\.]+f'. Slicing by floats prints a warning.

======

Backward Incompatible Changes:
- Arguments to viscid.field.wrap_field have changed
- typ arguments have been refactored to fldtype and crdtype
- add Field methods real, imag, astype

Changes:
- add support for calling `python -im viscid pylab`
- add viscid.field.arrays2field and viscid.coordinates.arrays2crds
- viscid.field now has empty, ones, zeros, empty_like, ones_like, zeros_like that act roughly the same as the numpy functions, except that these create fields

======

Changes:
- Ascii dataset field names are now "c[0-9]+" where the number is the column number
- Add a way to generate your own cubehelix colormaps
- Sphinx has no more warnings