Store count data outside of plot JSONs

[fedarko]

I think this should make the plots load a ton faster, as well as stave off crashing when you try to open the vega editor for the sample plot. (It's not like you can recompute log ratios within the Vega Editor, so there isn't any sense storing the count data inside the sample plot.)

This shouldn't require a ton of coding work. Just:

• change up the python script to output the count data to a separate variable from the sample plot
• change up the python testing utilities to get count data from that variable instead
• change up the JS to take the count data as an RRVDisplay parameter, and set this.feature_cts according to that
• change up the JS tests to set the count data accordingly as well

[fedarko]

We might be able to take this a step farther, and actually load the count data (and compute log ratios, or just get the desired counts) entirely in a web worker. This would prevent the costs of loading that into memory from hitting the page, ideally.