fix(channelserver): handle ignored DB errors and cache userID on session

Silently ignored DB errors in handlers could cause data loss (frontier
point transactions completing without DB writes), reward duplication
(stamp exchange granting items on failed UPDATE), and crashes (tower
mission page=0 causing index-out-of-bounds). House access state
defaulting to 0 on DB failure also bypassed all access controls.

HIGH risk fixes:
- frontier point buy/sell now fails with ACK on DB error
- stamp exchange/stampcard abort on failed UPDATE
- guild meal INSERT returns fail ACK instead of orphaned ID 0
- mercenary/airou creation aborts on failed sequence nextval

MEDIUM risk fixes:
- tower mission page clamped to >= 1 preventing array underflow
- tower RP donation returns early on failed guild state read
- house state defaults to 2 (password-protected) on DB failure
- playtime read failure logged instead of silently resetting RP

Also cache userID on Session at login time, eliminating ~25 redundant
subqueries of the form WHERE u.id=(SELECT c.user_id FROM characters
c WHERE c.id=$1) across shop, gacha, command, and distitem handlers.

server/channelserver/sys_session.go

@@ -45,6 +45,7 @@ type Session struct {
 	stagePass        string // Temporary storage
 	prevGuildID      uint32 // Stores the last GuildID used in InfoGuild
 	charID           uint32
+	userID           uint32
 	logKey           []byte
 	sessionStart     int64
 	courses          []mhfcourse.Course
@@ -352,7 +353,7 @@ func (s *Session) GetSemaphoreID() uint32 {
 
 func (s *Session) isOp() bool {
 	var op bool
-	err := s.server.db.QueryRow(`SELECT op FROM users u WHERE u.id=(SELECT c.user_id FROM characters c WHERE c.id=$1)`, s.charID).Scan(&op)
+	err := s.server.db.QueryRow(`SELECT op FROM users WHERE id=$1`, s.userID).Scan(&op)
 	if err == nil && op {
 		return true
 	}