Noticeable CPU usage when net._http_response has high count

[steve-chavez]

This will be fixed on #68. Opening the issue to make the bug clear.

---

When having

postgres=# select count(*) from net._http_response;
-[ RECORD 1 ]
count | 50001

postgres=# select count(*) from net.http_request_queue;
-[ RECORD 1 ]
count | 0

Usage stays mostly at 2.6%CPU. Also notice the 6.7% MEM, which indicates there's a memory leak.

This is due to this DELETE:

pg_net/src/worker.c

Lines 442 to 445 in 70ea4cd

	char *sql = "DELETE FROM\n"
	" net._http_response\n"
	"WHERE\n"
	" clock_timestamp() - created > $1::interval";

clock_timestamp() generates a full table scan each time it runs.