User directory rebuild spends too long calculating progress

[reivilibre]

On matrix.org we have this applied:

-            txn.execute("SELECT COUNT(*) FROM " + TEMP_TABLE + "_rooms")
-            result = txn.fetchone()
-            assert result is not None
-            progress["remaining"] = result[0]
+            #txn.execute("SELECT COUNT(*) FROM " + TEMP_TABLE + "_rooms")
+            #result = txn.fetchone()
+            #assert result is not None
+            #progress["remaining"] = result[0]

because the SELECT COUNT(*) involves a table scan. We should either skip this or do something smarter (incrementally update number OR use the postgres stats tables for a cheap size estimate).

[reivilibre]

woops, this was fine and made it into #15435