Fixes for iron derivation for GreaterEqual, LessEqual, In

[kciesielski]

These fixes should address issues reported here: https://softwaremill.community/t/how-to-improve-open-api-documentation-for-iron-refined-types/396/9

1. LessEqual and GreaterEqual validators aren't resolved correctly if these predicates are combined with other predicates using Or or And. They would be handled by validatorForDescribedOr, so I added variants called validatorForDescribedOrGe and validatorForDescribedOrLe witch additional "witnesses" like isGe: P =:= GreaterEqual[Num] to redirect back to validatorForGreaterEqual and validatorForLessEqual.

2. Constraint In[(1, 2, 3)] was handled as a standard union type constraint, resulting in a list of validators, where each validator was a Validator.enumeration(List(int)) with a single value in the list. I have added some tweaks to fix this:

• constrains like In[(a, b, c, d)] are now represented as a single Validator.enumeration(List(a, b, c, d))
• raw strict equals in unions are handled as well, so constraint like StrictEquals[1] | StrictEquals[2] will also produce a single enumeration
• Mixed unions of StrictEquals and other constrains are also supported, as long as StrictEquals are the first constrains in the union, for example: StrictEquals[3] | StrictEquals[4] | GreaterEqual[23] will give a Validator.Enumeration(List(3, 4), _, _), Validator.Min(23, false))

[adamw]

Nice! :) Thanks