Refactor to use holoviews powered bar charts

[AjayThorve]

This PR refactors the bokeh charts implementation, and uses holoviews to generate the charts, while also using box-selects instead of range_sliders. This was done for 2 reasons, stability of holoviews api, which also is much more efficient, and to use streams + custom callbacks to link all charts with bar_charts.

Some of the other notable changes:

• Using holoviews to generate bokeh plots, allows us to str columns directly for bar charts
• Persistent non-selection view for bar_charts
• Got rid of redundant bokeh line chart, and now a single implementation exists under datashader/line, callable cuxfilter.charts.line
• removed preview as pyppeteer was unstable
• Refactor to remove redundant functions and classes

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[AjayThorve]

Lot of changes! Whats the quick rational for removing the bokeh line chart?

Updated the description,

Adding to it: The current implementation had two line charts, one under bokeh, and another under datashader. Removed one to avoid confusion, and bokeh line's implementation that cuxfilter had wasn't really accomplishing anything that bar charts weren't.

[AjayThorve]

/merge