First, connection pools do not avoid multiple connections to a database. What they do is limit the overall number of connections open at any given time and re-use them between parallel threads or processes. They may reduce the number of connections needed at one time, but only if some of the threads aren't doing any database work.

Second, you don't need one. Just fork processes for the number of connections you want, connect separately in each (after the fork) and go. That's the most efficient you can get with this.