Made hvplot API dynamic by accepting parameter/widget instances

[jsignell]

NEW: Replacing #279 and updating to accept widgets as well as parameters.

This PR makes it possible to build GUIs by passing parameters instances directly into the hvPlot methods. Combined with Panel this makes it very easy to build complex GUIs.

import panel as pn
import hvplot.pandas

from bokeh.sampledata.iris import flowers

cols = list(flowers.columns[:-1])
x = pn.widgets.Select(name='x', value='sepal_length', options=cols)
y = pn.widgets.Select(name='y', value='sepal_width', options=cols)
kind = pn.widgets.Select(name='kind', value='scatter', options=['bivariate', 'scatter'])
by_species = pn.widgets.Checkbox(name='By species')
color = pn.widgets.ColorPicker(value='#ff0000')

@pn.depends(by_species, color)
def by_species_fn(by_species, color):
    return 'species' if by_species else color

plot = flowers.hvplot(x=x, y=y, kind=kind, c=by_species_fn)

pn.Row(pn.WidgetBox(x, y, kind, color, by_species), plot)

This examples demonstrate how both parameters and functions that depend on some parameter(s) can be used to dynamically update a plot. In this case the by_species_fn returns either the 'species' column or the ColorPicker depending on whether the checkbox is ticked, making it simple to specify even complex behaviors.

Currently this re-renders the plot every time a parameter changes. A future exercise can be to optimize this by handling certain parameters internally using DynamicMaps.

• Clean up code
• Add docs
• Add tests

closes #325

[jsignell]

I am exposing this as well:

import panel as pn
import hvplot.pandas

from bokeh.sampledata.iris import flowers

flowers.hvplot.scatter(groupby='species', legend='top_right', 
                       widgets={'species': pn.widgets.DiscreteSlider})