Currently, the fact that an index is clustered is not dumped. Unfortunately
the only way of dumping this information is with a cluster statement itself.
One possible improvement would be to sort clustered indexes first and dump
them first, that way the cluster operation does not have to do so much
reindexing.
Example output:
--
-- TOC entry 7 (OID 17078)
-- Name: test_idx; Type: INDEX; Schema: public; Owner: chriskl
--
CREATE INDEX test_idx ON test USING btree (a);
CLUSTER test_idx ON test;