Token scoring

[anderdc]

use tree-sitter to parse github code content and calculate a 'truer' value of changed lines

• validators now fetch code contents, not just file patches, this is for creating ASTs and comparing old vs new nodes/tokens
• using old and new file contents, construct AST's and compare changed tokens
• structural and leaf code changes are calculated each with their own set of weights
• file types we don't support do not get scored (lines do count towards code density)
• code_density is token_score / total_lines
• base score is no longer static, gets multiplied by code_density
• no base score earned for PRs with < 5.0 token score, contribution bonus can still be earned
• contribution_bonus uses token_score
• tier configs now use token_score, need min token score AND a min token score for n unique repos per tier now
• PRs are no longer considered "low value" and disregarded for scoring. but since token score and code density play a big part in scoring, they will still have a low value score-wise
• left in low_value_pr code as deprecated code in case we flip back
• etc...

misc. updates

• remove db migrator and definitions of db tables