Conservation chaos run on ScyllaDB

The conservation audit (#128) held through a sim crash and a rolling restart on the in-process cluster (one memory store). Run the same scenario on ScyllaDB with node processes (M7-style, #89/#90), so the logged CQL batches are proven under a real crash:

UO_RAID_WINDOW=always cargo run --release -p uo-loadtest -- scenario --store scylla://127.0.0.1:9042 \
  --sims 3 --gateways 2 --bots 500 --mix miner=20,lumberjack=10,fisher=5,hunter=5,hauler=10,raider=5,guard=3 \
  --chaos "60:kill:sim-2,75:start:sim-2,110:rolling" --map 512

Done when report.json › conservation shows every audit balanced, and a backup taken during the run restores (uo-server store restore) on a second keyspace. Also rehearse the restore on the production Scylla once game-hosting exists.