Add Collection.plot() function

[mnlevy1981]

As a first pass, I allow Collections to contain matplotlib.figure.Figure
objects. Collection.plot() requires a variable argument (for now a string, but
perhaps a list of variables to plot would be better?) and also allows the user
to pass isel_dict to help reduce dimensions. There is a trello card about
introducing sel / isel functionality, and once that's done the isel_dict
argument can be removed.

I like the idea of being able to super-impose line plots (time series of scalar
quantities, or vertical profiles) in a single figure, but I have not made any
progress on that front yet.

I also should add a test for this function.

Fixes #13

[mnlevy1981]

I was on the fence about what Collection.plot() should return; maybe instead of an Collection, a regular old dictionary would suffice? Letting a Collection contain figure.Figure objects is likely going to complicate a lot of the other functions provided by the class, and really all I want is a way to access the figures later. For example,

ds = xr.tutorial.open_dataset('rasm')
dsa = xr.tutorial.open_dataset('air_temperature')
collection = xc.Collection({'rasm': ds, 'air_temp': dsa, 'air_temp_to_Tair': dsa.rename({'air': 'Tair'})})
plot_col = collection.plot('Tair', {'time': 0})

plot_col['rasm'].savefig('Tair.png')