Redo the plotting routines using a more compact layout, better suited for use with bumps webview.

This code is not used by SasView

Notes:

(1) Include both matplotlib and plotly backends.

The plotly backend isn't enabled because the bumps interface to plotly in model is not yet stable. See bumps/bumps#342

(2) plotly heatmaps do not yet display correctly in bumps

To see the plots in matplotlib and plotly type the following to load and display the plots without bumps:

python example/fit2d.py

Now switch to the fix-improve-mpl-support branch in bumps, uncomment def plotly(...) in bumps_model.Experiment.plot, then run

bumps example/fit2d.py

The plotly 2d plots are still ugly. Figure out how to shrink the graph area to just the data

(3) The limits parameter for plot_theory doesn't work for q^4 plots

(4) matplotlib log data error plots do not work with webview

This is an issue with the mpld3 library used to render them: mpld3/mpld3#488

(5) Should be sharing the 2D rebinning code with sasview

(6) The algorithm for select Δqx and Δqy when converting from a bag of (qx, qy, I(qx, qy)) values to a 2D image should be improved. Pixels should be roughly square; instead of assuming a square aspect ratio it should use the qx and qy extent ratio. The algorithm is assuming that (qx, qy) is relatively uniform throughout bounding box so it is creating a grid with approximately one measurement per pixel. This fails if the dataset has significant masking, or if it combines measurements with significant overlap. Either find a better algorithm as part of this PR or create a separate ticket asking to improve it.

(7) Although these changes mirror some of those in the sasview plotting refactor (SasView/sasview#2940 and SasView/sasview#3111) it doesn't look like higher level code sharing will be possible.