Area weighted regridding

[esc24]

The PR adds area weighted regridding to iris.experimental.regrid.

It is not the finished product:

1. More testing is needed,
2. I need to handle aux factories so the surface altitude is regridded using area weighting.
3. Removal of debug print statements.

I have put it up so @bblay can begin the review process.

Note the travis failures are due to an expectation that is_contiguous will be modified (#459), partial aux factory support and a dependency on numpy 1.7 to support a tuple as the axis argument.

[bblay]

Helpful hint: To present a PR without the risk of accidental merge, pull it to your own master...You have got a master, haven't you? 😉

[esc24]

Thanks @bblay, but I trust the failing tests and 23 commits would prevent accidental merging 😉.

[bblay]

rogue word "coordinate"?

[bblay]

Out of scope really as it's not your code, but _create_cube doesn't describe the conditions in which it can trigger a KeyError. An example would be really helpful in understanding this.

[bblay]

Great job @esc24.
I have reviewed the main code and will review the tests and play with it next.

[bblay]

The code is set up to handle the x or y coord being scalar, as we'd expect from a cross-section. Please add tests for x and y cross sections (consider adding one for a single-column cube too).

[bblay]

Please add a tests for non-contiguous x and y bounds.

[bblay]

Please add a test for incompatible coords (i.e name, units or attribute, as detected in Coord.is_compatible)

[bblay]

Extending the request for cross-section tests, please include attempts to regrid between a cross section and a horizontal grid.

[bblay]

Please add tests for missing x and y coords.

[bblay]

This comment about a spherical effect seems wrong here, we just disabled spherical calculations.

[esc24]

Removed.