Better detect unreferenced definitions in the linter

[jviotti]

Signed-off-by: Juan Cruz Viotti jv@jviotti.com

[augmentcode]

🤖 Augment PR Summary

Summary: This PR re-enables and improves detection of unreferenced schema definitions in the linter by updating the vendored sourcemeta/core and switching the transformer framing to reference-oriented analysis.

Changes:

• Updates the vendored core dependency to a newer revision.
• Re-enables the orphan_definitions lint rule in src/command_lint.cc.
• Refactors OrphanDefinitions to identify orphaned entries in $defs/definitions using resolved reference destinations (rather than instance-location tracking).
• Changes SchemaTransformer framing from SchemaFrame::Mode::Instances to Mode::References to support reference-based lint rules.
• Adds JSON Pointer helpers (Pointer::starts_with(...)) and introduces a new mangle(PointerTemplate, prefix) utility.
• Adds a new linter rule unsatisfiable_in_place_applicator_type and updates lint rule list tests/counts accordingly.

Technical Notes: The linter rule inventory grows (rule list counts updated), and orphan detection now explicitly considers references coming from outside the definitions containers.

_{🤖 Was this summary useful? React with 👍 or 👎}

[augmentcode]

Review completed. 2 suggestions posted.

Comment augment review to trigger a new review at any time.

[augmentcode]

TokenVisitor::operator()(const Pointer::Token&) unconditionally calls token.to_property(), but PointerTemplate can contain index tokens (e.g. { "foo", 1, "bar" }), which will trip the to_property() assert at runtime. It seems worth handling token.is_index() (or explicitly constraining mangle to property-only templates) to avoid crashes.

_{🤖 Was this useful? React with 👍 or 👎}

[augmentcode]

encode_prefix() doesn’t appear to apply the same “reserved at segment start” handling described in the public doc comment for mangle (e.g. distinguishing x/X/X.. escape prefixes), which can lead to prefix collisions. If mangle is meant to be collision-free for arbitrary prefixes, it may be worth aligning prefix encoding with the documented rules (or documenting allowed prefix constraints).

_{🤖 Was this useful? React with 👍 or 👎}