Axiom types refactor

[dkubb]

This branch will include a refactoring of axiom to use axiom-types.

• [x] Remove unnecessary DEFAULT_SIZE constants since the types use the defaults from axiom-types anyway
• [x] Move static #type methods to become class methods, and have #type delegate to it
• [x] Remove Axiom::Attribute::Date#range
• [x] Remove Axiom::Attribute::DateTime#range
• [x] Remove Axiom::Attribute::Time#range
• [x] Rename options to Axiom::Attribute::String constructor to be :minimum_length and :maximum_length instead of :min_length and :max_length
• [x] Update Axiom::Attribute::String constructor to only create a custom type only if the :minimum_length or :maximum_length options are specified
• [x] Update Axiom::Attribute::Numeric constructor to only create a custom type if a :size option is specified
• [x] Remove the @options ivar from Axiom::Attribute. Descendants should pull what they need from the options and store an explicit ivar
• [x] Add a method to axiom-types where given two types we can find the most common ancestor, for example: type.common_ancestor(other).
• [x] Use Axiom::Types::Type#common_ancestor to return the best type from Axiom::Function::Numeric::Binary#type
• [x] Audit diff to make sure @example YARD docs, and others, were not accidentally removed during the conversion
• [x] Ready for review
• [x] Squash commits

[dkubb]

@cored ok I think I have something working that passes the tests.

There's some performance issues with this though, the specs run much slower than before and I think it's in Attribute.infer_type, which I left a comment on.

I also think we should audit the Attribute subclasses and make sure our quick patches didn't miss anything. Ideally we want to remove as much as possible from each Attribute subclass, and delegate things to the type, when possible. I've already left comments hinting in this direction, but I'm sure there's more we can do.

I am going to do a pass over the diff and look at everything we did to make sure there's no extraneous stuff, and make sure it's optimal. Feel free to do the same if you want, the more of us looking at this the better.

[dkubb]

@cored I just added the --warnings flag to rspec and saw a ton of warnings in axiom-types. I will get those fixed up today, but I wanted to give you the heads up so you didn't think it was something you did.

[dkubb]

@cored on the TODO list, I have some notes about removing #range and delegating #minimum and #maximum to the type. I wonder if those are even necessary? Maybe we should just start by removing the #range method and not worry about delegating those other methods until a concrete need appears? wdyt?

[dkubb]

@cored I also noticed that we have things like #min_length in String, and it only seems to be used for equalizer's benefit. If we changed Axiom::Attribute itself to equalize on the name, type and required?, then we could eliminate these methods.

Previously all these methods were needed in Attribute because it was handling constraints and validations. Now all that is handled by axiom-types. Maybe someday we'll need a way to get to the type constraints through the attribute, but I would rather remove code until that is totally required.

[dkubb]

@cored after thinking about this for a while, I think it is probably unnecessary to completely eliminate the Axiom::Attribute subclasses. I think anything we do to create an intermediary structure to link together methods and types will end up being far more complex than Axiom::Attribute subclasses.

However, I do think we should continue with removing functionality that is duplicated between them. Please note this does not change anything we've done up to this point, and does not affect anything still to be done in this PR. All it means is that once the todo notes in this PR are complete and this code is merged into master, we're essentially done with the integration.