feat(clickhouse): integrate goose migrations and analytics repository

Add a separate ClickHouse OLAP pipeline with goose/clickhouse DDL as the
sole schema source, model/analytics for ORM mapping, and repository/analytics
for reads (ChDB/GORM) and batch writes (ChConn). Refactor risk_control and
admin/logs to use the repository layer instead of inline SQL. Remove the
manual support-files DDL and document the workflow in database-migration skill.

internal/repository/analytics/access_log_writer.go

@@ -0,0 +1,49 @@
+// Copyright 2026 Arctel.net
+// SPDX-License-Identifier: Apache-2.0
+
+package analytics
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/Rain-kl/Wavelet/internal/db"
+	analyticsmodel "github.com/Rain-kl/Wavelet/internal/model/analytics"
+)
+
+// BatchInsert writes access logs to ClickHouse using the native batch API.
+func BatchInsert(ctx context.Context, logs []analyticsmodel.UserAccessLog) error {
+	if len(logs) == 0 {
+		return nil
+	}
+	if db.ChConn == nil {
+		return fmt.Errorf("clickhouse connection is not initialized")
+	}
+
+	batch, err := db.ChConn.PrepareBatch(ctx, analyticsmodel.UserAccessLog{}.BatchInsertSQL())
+	if err != nil {
+		return fmt.Errorf("prepare clickhouse batch: %w", err)
+	}
+
+	for _, logItem := range logs {
+		if err := batch.Append(
+			logItem.ID,
+			logItem.UserID,
+			logItem.Path,
+			logItem.Method,
+			logItem.IP,
+			logItem.UserAgent,
+			logItem.Headers,
+			logItem.Status,
+			logItem.Latency,
+			logItem.CreatedAt,
+		); err != nil {
+			return fmt.Errorf("append access log to batch: %w", err)
+		}
+	}
+
+	if err := batch.Send(); err != nil {
+		return fmt.Errorf("send clickhouse batch: %w", err)
+	}
+	return nil
+}