Linux 7.0: explicitly set setlease handler to kernel implementation

The upcoming 7.0 kernel will no longer fall back to generic_setlease(),
instead returning EINVAL if .setlease is NULL. So, we set it explicitly.

To ensure that we catch any future kernel change, adds a sanity test for
F_SETLEASE and F_GETLEASE too. Since this is a Linux-specific test,
also a small adjustment to the test runner to allow OS-specific helper
programs.

Sponsored-by: TrueNAS
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Rob Norris <rob.norris@truenas.com>
Closes #18215

module/os/linux/zfs/zpl_file.c

@@ -43,6 +43,9 @@
 #ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO
 #include <linux/writeback.h>
 #endif
+#ifdef HAVE_FILELOCK_HEADER
+#include <linux/filelock.h>
+#endif
 
 /*
  * When using fallocate(2) to preallocate space, inflate the requested
@@ -1242,6 +1245,7 @@ const struct file_operations zpl_file_operations = {
 	.mmap		= zpl_mmap,
 	.fsync		= zpl_fsync,
 	.fallocate	= zpl_fallocate,
+	.setlease	= generic_setlease,
 	.copy_file_range	= zpl_copy_file_range,
 #ifdef HAVE_VFS_CLONE_FILE_RANGE
 	.clone_file_range	= zpl_clone_file_range,