Update comparables vignette terminology + add section re: comparable stability

[dfsnow]

After sharing the comparables finding vignette with some folks, we received some good suggestions for improvement:

• Change the term "target property" to "property of interest" since the former has a specific meaning in machine-learning contexts that doesn't apply here.
• Add a final section to the vignette testing the stability of comparable scores across different hyperparameter settings and numbers of iterations. The idea here being, any comparables we derive using this method should be robust to different model structures, given enough iterations.
• Look into random forest proximity scores and compare to this comps method.