3bef6b38c7
The `materialized` INTEGER column was redundant: create()/update() already keep a lazy session in memory and write no row, so a `sessions` row is written only by the first append. Its EXISTENCE is the materialization signal — has()/list() now report exactly the sessions that have a row, matching the JSONL backend's "file exists ⇔ materialized". The column only existed to force has()/list() to FALSE for an all-tail crash (a partial first turn, zero committed events). That actually DIVERGED from the JSONL backend, whose file (and thus has()=true) survives a first append that never reached turn/end. Removing the column drops that special case: an all-tail session keeps its row and stays present, the same as JSONL. The orphaned tail rows are still removed by the deferred truncation-repair on the next append, and load() stays non-mutating. Also: add a TODO to route through a cordis db service if one is adopted, and correct the README's Node-version framing to the repo's engines (>=24).