test(format/uri): add strict percent-encoding syntax and non-ASCII character cases

[AcEKaycgR]

Applies to: draft-v1, draft-04, draft-06, draft-07, draft/2019-09, draft/2020-12

Adds 12 new cases validating strict percent-encoding syntax triplets, caret/hash delimiters, and non-ASCII character constraints in the uri format.

Added:

• Incomplete percent-encoding % + 1 digit in query (invalid)
• Invalid hex digit %G0 in query (invalid)
• Lone percent % in query (invalid)
• Raw non-ASCII characters in query (invalid)
• Caret ^ in query (invalid)
• Incomplete percent-encoding % + 1 digit in fragment (invalid)
• Invalid hex digit %G0 in fragment (invalid)
• Lone percent % in fragment (invalid)
• Second hash # in fragment (invalid)
• Caret ^ in fragment (invalid)
• Raw non-ASCII characters in fragment (invalid)
• Valid percent-encoding in host reg-name (valid)

Ecosystem Impact (Triangulation):

Under Bowtie verification against 8 active implementations, these cases successfully caught live compliance issues:

1. Go & PHP Bugs Caught: Both engines incorrectly accept incomplete percent-encoding ("%2"), invalid hex ("%G0"), and lone percents ("%") in queries and fragments.
2. Go Bug Caught: go-gojsonschema incorrectly accepts raw non-ASCII é and carets in queries and fragments.
3. Go Bug Caught: go-gojsonschema incorrectly rejects valid percent-encoding in host reg-names ("http://ex%61mple.com").

@jviotti @jdesrosiers @karenetheridge Ready for review.

[jviotti]

Interesting. I would have said some tests are very similar to others around percentage encoding, but I think we should accept them as-is if you managed to prove existing implementations break with them.