fix gcc compile floating-point computation program when using option -march=native

When compile UnixBench, using command as below:
	gcc -o pgms/float -Wall -pedantic -O0 -march=native -mtune=native -I ./src -DTIME -Ddatum=float src/arith.c
It will have an error:
	src/arith.c: In function ‘dumb_stuff’:
	src/arith.c:89:5: error: ‘+nofp’ feature modifier is incompatible with floating-point code

The root cause is that gcc will read /proc/cpuinfo, get supported features from the line begin with 'Features'.
But tk4 haven't 'Features', it using 'flags' instead, which causing gcc can't get the supported features.

The correct usage is "Features:" always, just a buggy patch change it to
'flags' stupidly. Now change it back.

Signed-off-by: Jianping Liu <frankjpliu@tencent.com>
Signed-off-by: Alex Shi <alexsshi@tencent.com>

arch/arm64/kernel/cpuinfo.c

@@ -171,7 +171,7 @@ static int c_show(struct seq_file *m, void *v)
 		 * rather than attempting to parse this, but there's a body of
 		 * software which does already (at least for 32-bit).
 		 */
-		seq_puts(m, "flags\t\t:");
+		seq_puts(m, "Features\t:");
 		if (compat) {
 #ifdef CONFIG_COMPAT
 			for (j = 0; compat_hwcap_str[j]; j++)