Add batch eval and card-skip A/B scripts

eval_batch.sh: N back-to-back sessions with per-session summaries.
ab_card_skip.sh: A/B the jev vs combined card-reward skip policy and
compare deck size and progress from the game's run history.

eval_batch.sh

@@ -0,0 +1,25 @@
+#!/usr/bin/env bash
+# Run N sessions back to back and collect the resulting run records.
+# The bot handles menu/character-select navigation itself, so no setup is needed.
+set -u
+
+# Fail fast if the bot cannot start a run (e.g. a pending Timeline epoch).
+if ! python3 run.py --steps 1 >/tmp/preflight.log 2>&1; then
+  cat /tmp/preflight.log
+  exit 2
+fi
+SESSIONS="${1:-4}"
+STEPS="${2:-600}"
+for i in $(seq 1 "$SESSIONS"); do
+  echo "=== session $i/$SESSIONS ==="
+  python3 run.py --steps "$STEPS" --pause 0.3 > "/tmp/eval_${i}.log" 2>&1
+  echo "  exit=$? takes=$(grep -c 'select_card_reward' "/tmp/eval_${i}.log") skips=$(grep -c 'skip_card_reward' "/tmp/eval_${i}.log")"
+  python3 - <<'PY'
+import sts2
+try:
+    d = sts2.state(); p = d.get('player') or {}
+    print(f"  now: {d.get('state_type')} run={d.get('run')} hp={p.get('hp')}/{p.get('max_hp')}")
+except Exception as e:
+    print(f"  state read failed: {e}")
+PY
+done