feat(catalog): add MoltSets enrichment provider

[shehjad-dev]

Why

Add MoltSets as a credentialed people and company enrichment provider while preserving treg's faithful-relay and money invariants.

What changed

• Added 17 direct MoltSets catalog tools with request/response examples.
• Enabled 9 fixed-cost, single-result tools for the shared plan at $0.01 per call.
• Kept 8 variable-result, batch, or dual-meter tools BYOK-only.
• Added four routing adapters for name-to-email and person enrichment by name, email, or LinkedIn URL.
• Added account-based capacity collection and policy handling for the rolling API quota.
• Added MoltSets to environment-key discovery and provider metadata.
• Added the official MoltSets logo.
• Added provider, pricing, routing, capacity, environment, and safety tests.
• Added the MoltSets architecture fragment and regenerated the context index/map and five treg skill mirrors.

Shared-plan boundary

Shared-plan tools:

• people.email.find
• people.email.find.name
• people.email.find.business
• people.email.find.personal
• people.email.find.personal-best
• people.enrich.name
• people.enrich.email
• people.enrich.linkedin
• linkedin.profile.from-email

BYOK-only tools:

• people.phone.find
• people.search
• companies.search
• companies.identify.ip
• linkedin.profile.search
• people.audiences.hashes
• people.audiences.maid
• people.audiences.sha256

The shared-plan price is $0.01 per ordinary call. MoltSets lists a $27/month plan, so the direct-plan break-even is 2,700 ordinary calls per month. Phone lookup remains BYOK-only because it consumes both an enrichment record and a phone token; its documented upper-bound economics are $0.51 per result.

Settlement uses the existing generic status == "ok" acknowledgement. No provider-specific relay or money branch was added.

Live verification evidence

Budget used: 49 API requests out of the 50-request cap, and 2 phone tokens out of the 2-token cap.

Observed against the funded test account:

• Invalid authentication returned 401.
• Account, billing, and usage probes returned 200 and did not consume enrichment records.
• Invalid request validation returned 422 and did not consume a record.
• An ordinary successful lookup decremented one enrichment record.
• An ordinary miss did not decrement a record.
• A search page returned successfully and decremented one search record.
• A successful phone lookup decremented one enrichment record and one phone token.
• All 16 non-phone paths were exercised with synthetic misses and returned the documented 200/not-found envelope.
• Phone success and miss behavior were both verified.
• Free-plan and $27-plan account shapes were observed.

Account, billing, and usage endpoints are intentionally internal capacity probes rather than direct catalog tools.

Verification

• tests/test_moltsets.py: 20 passed
• Focused provider/integration set: 25 passed
• Relevant catalog, money, capacity, and routing set: 185 passed
• Catalog validation: 103 provider files, 3,361 endpoints, 0 errors, 0 warnings
• Import contracts: 14 kept, 0 broken
• Five generated treg skill mirrors: clean
• Context drift review: MoltSets and affected shared fragments updated
• git diff --check: clean
• Full parallel suite with local sockets enabled and TREG_OVERFLOW_MODE=off: 4,100 passed, 8 skipped, 92 failed. Every failure is confined to the pre-existing missing agent/use-case pages and dependent SEO assertions; there are no MoltSets, catalog, routing, capacity, money, or architecture failures.

Operations

This PR does not enable MoltSets in production, change private deployment configuration, purchase anything, or enable auto-top-up. After review, production enablement requires the MoltSets platform key and allow-list entry to be configured through the private operational repository.