Add adventure plugin, holdem CFR fixes, and wordle plugin

Adventure: Complete v1 daily idle RPG with DM-driven gameplay, equipment
shop, treasure system, TwinBee NPC, streak/grudge/party mechanics,
flavor text, and scheduled morning/evening/midnight tickers.

Holdem CFR: Fix three critical training bugs (fold not forfeiting pot,
free calls after raises, training/runtime key mismatch). Add performance
optimizations (preflop lookup table, zero-alloc equity, integer keys,
raise cap, regret pruning). Enrich abstraction with 12 equity buckets,
board texture dimension, and 6-char action history. Replace validation
with proper multi-street simulation.

Also includes wordle plugin, holdem seed tooling, and schema additions.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

cmd/holdem-seed/main.go

@@ -0,0 +1,36 @@
+//go:build training
+
+package main
+
+import (
+	"fmt"
+	"log/slog"
+	"os"
+
+	"gogobee/internal/plugin"
+)
+
+// Creates a minimal seed policy so the bot can start without a trained policy.
+// Run: go run -tags training ./cmd/holdem-seed/
+func main() {
+	slog.SetDefault(slog.New(slog.NewTextHandler(os.Stdout, &slog.HandlerOptions{Level: slog.LevelInfo})))
+
+	data := &plugin.CFRData{
+		Regrets:  make(plugin.RegretTable),
+		Strategy: make(plugin.RegretTable),
+		Meta: plugin.CFRTrainingMeta{
+			Iterations: 0,
+			Seed:       42,
+			Date:       "seed-policy",
+		},
+	}
+
+	os.MkdirAll("data", 0o755)
+	if err := plugin.SaveCFRData("data/policy.gob", data); err != nil {
+		fmt.Fprintf(os.Stderr, "Error: %v\n", err)
+		os.Exit(1)
+	}
+
+	slog.Info("Seed policy created", "path", "data/policy.gob")
+	slog.Info("Train a real policy with: go run -tags training ./cmd/holdem-train/ --iterations 1000000 --workers 8")
+}