Fix setting next multixid's offset at offset wraparound

In commit 789d65364c, we started updating the next multixid's offset
too when recording a multixid, so that it can always be used to
calculate the number of members. I got it wrong at offset wraparound:
we need to skip over offset 0. Fix that.

Discussion: https://www.postgresql.org/message-id/d9996478-389a-4340-8735-bfad456b313c@iki.fi
Backpatch-through: 14

src/backend/access/transam/multixact.c

@@ -875,6 +875,7 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
 	int			next_pageno;
 	int			next_entryno;
 	MultiXactOffset *next_offptr;
+	MultiXactOffset next_offset;
 
 	LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
 
@@ -961,11 +962,15 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
 		next_offptr += next_entryno;
 	}
 
-	if (*next_offptr != offset + nmembers)
+	/* Like in GetNewMultiXactId(), skip over offset 0 */
+	next_offset = offset + nmembers;
+	if (next_offset == 0)
+		next_offset = 1;
+	if (*next_offptr != next_offset)
 	{
 		/* should already be set to the correct value, or not at all */
 		Assert(*next_offptr == 0);
-		*next_offptr = offset + nmembers;
+		*next_offptr = next_offset;
 		MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
 	}