Support Xarray Plotting Routines

[philipc2]

Closes #759

Overview

• Adds support to call Xarray plotting methods through the UxDataArray and UxDataset plotting accessors

Expected Usage

import uxarray as ux

grid_path = "/path/to/grid.nc"
data_path = "/path/to/data.nc"

uxds = ux.open_dataset(grid_path, data_path)

# call Xarray Plotting Methods
uxds['var'].plot.hist()
uxds['var'].plot.line()
# and so on

[philipc2]

Examples

uxds['v1'][0][0].plot.hist(bins=60, figsize=(10, 5))

uxds['v1'][0].mean("n_node").plot.line("o--", figsize=(10, 5))

[philipc2]

Update the .plot accessor docstring to mention the use of Xarray functionality.

[philipc2]

I have fixed the matplotlib backend issue we were encountering. Previously, if we would call backend='matplotlib' for any of our HoloViz functionality, it would set the backend used in the matplotlib package to agg, which differs from the default one in xarray, which is set to "module://matplotlib_inline.backend_inline"

With the changes, we the xarray methods no longer have any issues when backend='matplotlib' is set for any of our HoloViz functionality

# holoviews with a matplotlib backend
uxds['t2m'].plot(backend='matplotlib')

# xarray method, would not render before, but with the change it works
uxds['t2m'].plot.hist()