spl: make 'spl_panic_halt' working for all cases

The default behavior where the serious ZFS errors cause FS thread to
stuck is very bad for some production scenario.

In some production scenarios (Linux), it is recommended to make real
kernel PANIC, where system can be rebooted by watchdog or kernel itself.
This patch enables coherent handling of spl_panic_halt parameter.

Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Authored-by: Wojciech Nizinski <w.nizinski@grinn-global.com>
Signed-off-by: szubersk <szuberskidamian@gmail.com>
Closes #12120
Closes #13109

module/os/linux/spl/spl-err.c

@@ -100,6 +100,9 @@ vcmn_err(int ce, const char *fmt, va_list ap)
 		break;
 	case CE_PANIC:
 		printk(KERN_EMERG "PANIC: %s\n", msg);
+		if (spl_panic_halt)
+			panic("%s", msg);
+
 		spl_dumpstack();
 
 		/* Halt the thread to facilitate further debugging */