[judge] Algorithm revamp

[sumanthratna]

from HackTJ 7.5 feedback from judges:

• Hung on to the project that is overall best—if a project is gone, it's gone
• Didn’t like single-elimination system: can’t give credit to good projects that had different strengths
• "anti-bias" is cause for concern (I understand—but I think an explanation of Crowd-BT would mitigate this concern?)
• old system, where you rank on a rubric, was better because of the above

ways to fix the above:

• don't do an "overall" decision (?)
• use a rubric (i.e., more than just 3 subcategories)
  • if we get rid of overall, aggregate the scoring from a single project into a single overall score to use for information gain
  • a bit more like the previous system (more nuance in voting) but doesn't come with the downside of "choose from 1 to 10"
• allow re-making a comparison with an old project (but the same comparison shouldn't be made, e.g., current can be repeated but the same prev and current shouldn't)