Use existing spec-coerce library for coercions

[WhittlesJr]

Here's my proposal to resolve #1 .

The spec-coerce library seems to me to be the most complete attempt at leveraging specs for data coercion. However, the interface is a bit more stringent than what cyrus-config has been accustomed to. Essentially, :spec should only be provided as a qualified keyword to a registered spec. Bare predicates (like int?) don't work in the way that cyrus-config has been using them in its tests and documentation.

So, in addition to deleting the parsing logic from cyrus-config and using sc/coerce! in its place, this PR updates the tests and documentation with the required usage thereof.

Personally, I see this as a positive change, as specs should almost always be registered to keywords using s/def anyway. So it's enforcing good code practice to an extent. Furthermore, spec-coerce's model of providing custom coercion decomplects the spec from the coercion:

(s/def ::json-coll-of-int (s/coll-of ::c/int)) ;; The spec
(sc/def ::json-coll-of-int json/parse-string)  ;; The coercion

vs

;; Spec and coercion conflated into one thing
(s/def ::json-coll-of-intn (from-custom-parser json/parse-string (s/coll-of int?))) 

This change is also pending a PR I made to spec-coerce.

[WhittlesJr]

Here's the aforementioned PR:

wilkerlucio/spec-coerce#26