RE: Improve eviction algorithm in ReorderBuffer

Dear Sawada-san,

> Agreed.
> 
> I think the patch is in good shape. I'll push the patch with the
> suggestion next week, barring any objections.

Thanks for working on this. Agreed it is committable.
Few minor comments:

```
+ * Either txn or change must be non-NULL at least. We update the memory
+ * counter of txn if it's non-NULL, otherwise change->txn.
```

IIUC no one checks the restriction. Should we add Assert() for it, e.g,:
Assert(txn || change)? 

```
+    /* make sure enough space for a new node */
...
+    /* make sure enough space for a new node */
```

Should be started with upper case?

Best Regards,
Hayato Kuroda
FUJITSU LIMITED
https://www.fujitsu.com/