feat: support minItems and maxItems

[bradzacher]

NOTE - this branch was based off of #241.
Github will show a number of unrelated commits until that PR is merged.
For a nicer reviewing experience, please only review commits after 5cb467a

---

Fixes #218

This PR:

• add support for generating tuple types with optional and non-optional elements if minItems / maxItems exists.
• handles having either minItems / maxItems greater than the length of the provided items.
• handles having maxItems less than the length of the provided items.
• converts array types to tuple types in the presence of minItems or maxItems.

[bcherny]

This looks super good overall. A few comments. Also, what do you think about updating generated output to be more safe? As-is:

type A = [string, string?, string?]
const a: A = ['a', undefined, 'c']

Instead, we could avoid accidentally allowing undefined through:

type B = [string] | [string, string] | [string, string, string]
const b: B = ['a', undefined, 'c']

[bradzacher]

I've addressed the other comments.

In regards to changing the types, I think it's a good idea.
I'll have a look at it tonight.

[bradzacher]

It makes the types verbose, but it adds a heap more type safety.
It also allows explicit assertion of the empty array state when no minItems is provided.

I'm a lot happier with the types this way!

[bcherny]

Almost there -- one last (minor) suggestion.

[bradzacher]

OOOOKay. Sorry, this PR keeps growing.

So in adding in a normalizer for minItems, I noticed that the normalizer wasn't traversing correctly. I added a naive normaliser (if (!('minitems' in schema)) { schema.minItems = 0 }), and it broke most of the tests because the normalization algorithm traversed every single object key equally.

I noticed that most of the normalizers had a check for stringify(schema) === stringify(rootSchema), i.e. only apply at the root, but this means that they don't get applied correctly for anything beyond that - i.e. definitions, allOf, anyOf, oneOf, not, items, properties....

So I added a traverser which will properly traverse the schema, which allows the normalizers to correctly apply to things which are actually schemas beyond the root schema.

This surfaced a few bugs which I fixed.

[bcherny]

Looks great. Thanks for the enhancement, and thanks for improving underlying infrastructure along the way, Brad!

[bcherny]

Published 7.0.0.