plotly visualizer as a standalone component

[yogeshVU]

Based on the discussion with @brollb

It would be great to have plotly visualizer(https://github.com/deepforge-dev/deepforge/tree/master/src/visualizers/widgets/PlotlyGraph) as a standalone component, which can be incorporated into different webgme projects for visualization of results. Following seems to be current approach of the plotly visualizer in deepforge

1. write an operation in python using matplotlib
2. we have a custom matplotlib backend (like how jupyter notebooks handle matplotlib) which converts the plot to a plotly JSON representation
3. the plotly JSON is printed to stdout and parsed by the plugin responsible for executing jobs
4. the plotly JSON is saved in a node in the model. The PlotlyGraph visualizer is the default visualizer for this node type
5. when the node is viewed, the PlotlyGraph visualizer fetches the JSON representation and passes it to plotly to visualize

[umesh-timalsina]

@yogeshVU, this is a great idea. There are couple of issues that might need to be hashed out:

1. Plotly's matploltib renderer uses mplexporter, which is not compatible with latest matplotlib versions and has come to its end of life in terms of maintenance. So, a package for converting matplotlib to plotly figures should/have to be maintained for full support.
2. Support for 3D plots is not good. Even with the customization we wrote, its still very nascent in terms of being feature complete.