types: fix regression for vendor-specific field in nvme_id_ns

Recent versions of nvme-cli have started reading vs from offset
392 instead of 384. Previous PRs coupled the use of nvme_id_ns
for use in namespace management (create_ns).  However, the NVMe
spec has a different structure for namespace management, with
only a subset of the fields allowed/shared and some additional
fields.

To fix this, remove lbstm and restore the proper length for vs
from nvme_id_ns.

I expect that create_ns() should fully switch over to the
nvme_ns_mgmt_host_sw_specified struct which seems aligned with
the spec (though has some newer fields than what is available in
the latest NVMe Command Set Specification Revision 1.0c). This
will have to be addressed separately in nvme-cli.

#307
linux-nvme/nvme-cli#1452

Signed-off-by: Chris Patterson <[email protected]>

Author: cjp256

src/nvme/types.h

@@ -1799,7 +1799,6 @@ enum nvme_lbaf_rp {
  *	      remains fixed throughout the life of the namespace and is
  *	      preserved across namespace and controller operations
  * @lbaf:     LBA Format, see &struct nvme_lbaf.
- * @lbstm:    Logical Block Storage Tag Mask for end-to-end protection
  * @vs:	      Vendor Specific
  */
 struct nvme_id_ns {
@@ -1843,8 +1842,7 @@ struct nvme_id_ns {
 	__u8			nguid[16];
 	__u8			eui64[8];
 	struct nvme_lbaf	lbaf[64];
-	__le64			lbstm;
-	__u8			vs[3704];
+	__u8			vs[3712];
 };
 
 /**