CI: Validate embedded schema examples to catch logic regressions

[ptiper]

Is your feature request related to a problem? Please describe.

Tracking issue to implement the CI safeguard proposed by @amithanda in Universal-Commerce-Protocol/ucp#429 (comment).

Our current ucp-schema lint job checks syntax but misses schema logic overlap for discovery and profile documents. To programmatically prevent regressions like the ones fixed in PR #429, we could validate embedded JSON Schema examples directly during the lint phase.

Refer to Amit's comment linked above for the complete technical deep dive and sample payloads.

Describe the solution you'd like

Tasks

• Update ucp-schema lint to validate all entries in examples arrays, failing the build on errors.
• Enforce a lint rule requiring embedded examples on primary discovery schemas.
• Backfill examples into the base schemas to cover standard profile variations.

Describe alternatives you've considered

No response

Additional context

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[ShuoRen-TT]

I checked the current repositories before starting on this. The first task appears to have already landed in ucp-schema#34 (commit 6c3fb58): ucp-schema lint now recursively validates every examples[] entry that can be compiled in its enclosing subschema, emits E008 for invalid values, and includes focused tests and README documentation.

That implementation validates examples when present, but it deliberately does not require schemas to contain examples. This leaves the two policy/content parts of this issue:

1. backfill representative examples into the agreed primary discovery schemas;
2. enforce that those specific schemas continue to provide examples.

I recommend doing these in that order. First establish a clean baseline by adding the agreed examples and letting the existing E008 check validate them. Then add a narrowly scoped missing-examples rule. A generic rule requiring examples on every JSON Schema or subschema would produce noise and would be difficult for ucp-schema to apply without repository-specific knowledge.

Could maintainers confirm three boundaries before I prepare a PR?

• Which files count as the primary discovery schemas: profile.json, ucp.json, capability.json, or another explicit set?
• Should the presence rule live in the generic ucp-schema linter, or in the UCP repository CI where the expected file set is known?
• Should a missing example be an error immediately, or initially a warning while examples are backfilled?