Common API (Data Source Abstraction)

This PR introduces vast refactoring. The main problem it addresses is that before, functions like `query` or `execute` accepted a Connection object only. And at some point, it has become inconvenient. For example, I have a Pool object, and every time I want to query something, first I have to fetch a connection and pass it into the function as follows: ``` (defn do-something [pool] (pool/with-conn [conn pool] (let [result (pg/query conn "select ...")]))) ``` Some people can mix a connection with a pool, and pass a connection into the function, which will lead to an error. This PR brings something that is common in next.jdbc: protocols. There is a single ISource protocol that gets extended by the following types: - clojure map (clojure config) - Connection - Pool - Config (java) - URI String Functions like `query` and `execute` accept not a connection any longer but any of these objects listed above. Also, now it's possible to create a connection or a pool not from a Clojure map, but from a URI as well. I hope the new docs cover my intentions. Also, long Clojure->Java building functions are moved to separate namespaces to save lines. cc @jgdavey so you're aware of these changes.