Optimize rendering of the study view KM charts

[alisman]

https://cbioportal.mskcc.org/study/summary?id=mskimpact

The two KM charts wildly tax the resources of the browser for large samples. Scripting time, measured in profiler, is cut by 60% when we remove them. And peak memory footprint is reduced from 600MB to 300MB.

Rendering points and allowing interactivity is probably not important for this view. Perhaps we can sample and remove tooltips?

[alisman]

@dippindots product team agreed to getting rid of the censoring crosses and then doing binning. The way binning should work is that we put events into bins by some configurable number of days. We can apply this optimization only when there are more than say, 10k samples.

[dippindots]

@alisman @inodb @jjgao
Comments from standup meeting:

• Remove censoring crosses for sure
• flooring instead of binning, e.g. 100.3 months and 100.9 months fall into the same data point generated for 100 - 101 months
• If we do flooring, modify the current tooltip to include new information: # patients censored # patients have event