@abtree/srp-refactor

Score the codebase for Single Responsibility violations, snapshot the initial scoring, pause for the human to pick one to tackle, refactor it in a bounded loop (re-scoring after every pass), do a multi-agent code review, and finish with a before-vs-after