Add support for datetime axis on Image

[philippjfr]

Adds support for datetime axes on Image, SheetCoordinateSystem, BoundingBox and Slice. The basics are all working including instantiation of an Image with datetime bounds and slicing and indexing. Will make it possible to generate datashaded images with datetime axes.

[jbednar]

Excellent!!!

[philippjfr]

Since this is the first modification to SheetCoordinateSystem and BoundingBox since it moved into HoloViews, one of @jbednar and @ceball can have the great honor of reviewing this PR.

[philippjfr]

While I think Image will correctly retain timezone information, I think datashade operations will discard them. I can't say I have a strong desire to tackle that issue right now.

[jbednar]

Since I got to it first, I will nominate @ceball. :-)

[philippjfr]

Needs lots of tests still of course.

[philippjfr]

Also worth noting that it's currently using microsecond precision, switching to nanosecond precision introduces issues that are hard to handle correctly.

[jbednar]

When this is done, maybe open an issue about the nanosecond precision issues and say that it's not something we expect to fix, but that if someone wants to use it for that purpose then here's what they would need to fix.

[philippjfr]

Okay, this is ready to review now. I've added a bunch of unit tests for the utilities and for the interfaces. Iris does not like datetime coordinates apparently so I've had to skip those tests.

[philippjfr]

This output:

HoloMap containing 28 items of type RGB
---------------------------------------

Key Dimensions: 
Date: Oct 25 2012 01:00 UTC...Nov 02 2012 16:00 UTC 
Layer Height (hPa): 850...1000 
Constant Dimensions: 
Frame: None...None 
Deep Dimensions: 
x: 0...400.0 
y: 0...350.0
R: 0...1 
G: 0...1 
B: 0...1 

changed to this:

HoloMap containing 28 items of type RGB
---------------------------------------

Key Dimensions: 
Date: Oct 25 2012 01:00 UTC...Nov 02 2012 16:00 UTC 
Layer Height (hPa): 850...1000 
Constant Dimensions: 
Frame: None...None 
Deep Dimensions: 
x: 0...400 
y: 0...350 
R: 0...1 
G: 0...1 
B: 0...1 

The latter is more correct so I've updated the test data but the push build isn't pulling it for some reason. It should still work so this can be merged after review.

[philippjfr]

I'm going to go ahead and merge this myself since other work depends on it and the best smoke test is likely using it a lot in action. @ceball promised he'd take a more in depth look even after merging.