task #043: squared tag affinity curve, remove feedback /3.0, raise weight

task #043: rebalance tag affinity and feedback scoring

add attribution benchmark with autoresearch sweep

feedback prior uses matched tags only, not all content tags

separate evaluation judgment from tag attribution (#030)

tune: make RRF anchors corpus-scaled and configurable

fix: evaluator and MCP feedback score real content tags, not phantom names (#030)

fix: sync CLI category updates with category levels

fix: preserve evaluation context through dedup merges (#031)

chore: add tracker validation

curation pipeline: quality gate + dedup + consolidation for extracted engrams

chore: clean up task and issue tracker

store session title from ai-title transcript entry in processed_sessions

per-turn extraction: min 20K char threshold + tests

coordinate per-turn and batch extraction to prevent double-processing