u-boot: v2026.04: helios64: fix OTP magic check (any-byte → all bytes)

Per CodeRabbit review on PR armbian#9675:
  is_valid_header() in board/sys_otp.c uses '||' between byte
  comparisons, returning 1 if *any* one byte of otp.magic matches
  the expected sequence. Practically every random byte pattern will
  trip at least one comparison, so the magic check is effectively
  meaningless — partially corrupt OTP data is treated as a valid
  Helios64 header.

This is a pre-existing bug in the Kobol vendor code, carried forward
unchanged through every U-Boot bump. We are migrating these board
files to v2026.04, so the right place to fix it.

Replace the chain with a single memcmp() against the expected magic:
  - removes the '||' vs '&&' confusion entirely;
  - reads as a normal magic-bytes check;
  - the constant lives in one place.

Author: iav

patch/u-boot/v2026.04/board_helios64/board/sys_otp.c

@@ -4,6 +4,7 @@
 #include <log.h>
 #include <env.h>
 #include <net.h>
+#include <string.h>
 #include <u-boot/crc.h>
 
 #include "sys_otp.h"
@@ -44,14 +45,9 @@ static inline int is_data_valid(void)
 
 static inline int is_valid_header(void)
 {
-	if ((otp.magic[0] == 'H') || (otp.magic[1] == '6') ||
-		(otp.magic[2] == '4') || (otp.magic[3] == 'N') ||
-		(otp.magic[4] == 'P') || (otp.magic[5] == 'V') ||
-		(otp.magic[6] == '1') || (otp.magic[7] == 0))
+	static const u8 expected_magic[8] = { 'H', '6', '4', 'N', 'P', 'V', '1', 0 };
 
-		return 1;
-
-	return 0;
+	return memcmp(otp.magic, expected_magic, sizeof(expected_magic)) == 0;
 }
 
 static int init_system_otp(int bus, int cs)