Passing axes to plot_density fails with several datasets

[astoeriko]

Describe the bug
I would like to pass existing axes to plot_density with the ax keyword. This works fine if I plot data from a single dataset. But I get an error when I try to plot several datasets at once (I think the error arises when it tries to produce a legend).
Plotting several datasets without providing axes also works fine.

To Reproduce

import arviz
import matplotlib.pyplot as plt

test_data = arviz.load_arviz_data('centered_eight')

# This works
fig, ax = plt.subplots(3, 3)
ax1 = arviz.plot_density(data=[test_data.posterior], var_names=['mu', 'theta'], ax=ax);
# This works as well
ax2 = arviz.plot_density(data=[test_data.prior, test_data.posterior], var_names=['mu', 'theta']);
# This does not work
fig3, ax3 = plt.subplots(3, 3)
arviz.plot_density(data=[test_data.prior, test_data.posterior], var_names=['mu', 'theta'], ax=ax3);

This is the error I get:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-133-7feead268671> in <module>
     11 # This does not work
     12 fig3, ax3 = plt.subplots(3, 3)
---> 13 arviz.plot_density(data=[test_data.prior, test_data.posterior], var_names=['mu', 'theta'], ax=ax3);

~/miniconda3/envs/sunode/lib/python3.7/site-packages/arviz/plots/densityplot.py in plot_density(data, group, data_labels, var_names, transform, credible_interval, point_estimate, colors, outline, hpd_markers, shade, bw, figsize, textsize, ax, backend, backend_kwargs, show)
    263     # TODO: Add backend kwargs
    264     plot = get_plotting_function("plot_density", "densityplot", backend)
--> 265     ax = plot(**plot_density_kwargs)
    266     return ax

~/miniconda3/envs/sunode/lib/python3.7/site-packages/arviz/plots/backends/matplotlib/densityplot.py in plot_density(ax, all_labels, to_plot, colors, bw, figsize, length_plotters, rows, cols, titlesize, xt_labelsize, linewidth, markersize, credible_interval, point_estimate, hpd_markers, outline, shade, n_data, data_labels, backend_kwargs, show)
     75     if n_data > 1:
     76         for m_idx, label in enumerate(data_labels):
---> 77             ax[0].plot([], label=label, c=colors[m_idx], markersize=markersize)
     78         ax[0].legend(fontsize=xt_labelsize)
     79 

AttributeError: 'numpy.ndarray' object has no attribute 'plot'

Additional context
ArviZ version: 0.7.0
matplotlib version: 3.2.1

[aloctavodia]

Thanks for reporting the issue. I will take a look at it.

[astoeriko]

Thanks for the fix. :-)