Node null check fix

[lewisjkl]

Background

• What do these changes do?
  Updates such that the node validator will not raise a false-positive related to a Document value being null.
• Why are they important?
  Currently causing an error to flag in one of our projects where a trait has a Map shape with a target type of Document.

Testing

• How did you test these changes?
  See unit test included.

Links

• Links to additional context, if necessary
• Issue #, if applicable (see here for a list of keywords to use for linking issues)

---

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[haydenbaker]

While this seems like a simple change, it would have some unintended consequences (i.e. type signature changes) in some smithy-based code-generators. Though I understand the intent, I also don't think we should allow for nulls in a list/map that isn't marked with sparse. Which leads me to your PR adding sparse trait to your shape, which is the right approach here.

[lewisjkl]

I can understand that position, I think maybe the documentation for Document could be updated to remove mention that "null" is a valid document value. As far as I can tell, "null" for Documents is not treated any differently than it is for a String/Int or any other primitive.

"Document types are represented by a JSON-like data model and can contain UTF-8 strings, arbitrary precision numbers, booleans, nulls, a list of these values, and a map of UTF-8 strings to these values."

That's the passage that made me infer document had a special handling of Null compared to other primitives (and the fact that this used to work, we've had this same test case for years now)