Dependent parameter shows up as required

Grape has the concept of dependent parameters, which is not supported in OpenAPI specs currently (although there is an issue for it). I have found that given something simple like:

params do
  optional :a, type: String
  given :a do
    requires :b, type: String
  end
end

will cause the generated OpenAPI spec to have the :b parameter listed as a required parameter for the request.

This functionality is not supported in OpenAPI specs so there isn't a "correct" answer, but I believe that the attribute dependent on another attribute should default to the other attribute's status. If attribute :a was required and there was a dependent attribute then obviously it would also be required and would show up as such. But if :a is optional, and :b is only required when :a is present, then it seems that :b is also optional since a request with neither :a nor :b is still valid.

@LeFnord, is this something you guys would be receptive to as a change? I know it isn't following OpenAPI specs since this lies outside of it, but it seems correct to me. If you are open to it, I can make a PR to change the functionality.

Reproduce the issue with the Grape params block using optional :a and dependent required :b, then inspect how the generated OpenAPI spec determines parameter requiredness. Add coverage for optional and required parent parameters; done when :b is not marked required for optional :a but is marked required when :a is required.