[wberri13]

What this does

Closes #1287. Makes podcast episodes and transcripts discoverable through the existing exa.web.search tool and documents the native Exa domain filter, transcript-content request, and variable pricing. The general web-search example remains intact.

How it was tested

• The full suite passed on this commit: python3 -m uv run --frozen --with pytest-xdist pytest -n auto -q --disable-warnings — 6,878 passed, 132 skipped.
• python3 -m uv run --frozen python scripts/catalog_validate.py checked 4,089 endpoints with 0 errors.
• New regression tests cover the two podcast discovery queries, the native request body, and the agent-facing strong-result path. General Exa web search still resolves.
• The ticket reports a live Treg smoke test: HTTP 200, two podcast episode results, timestamped speaker text, and $0.007 charged with text capped at 1,000 characters per result. This PR's response excerpt is illustrative from Exa's public podcast page, because the raw smoke-test response was not available.

Checklist

• uv run --with pytest-xdist pytest -n auto -q passes locally (run through python3 -m uv with --frozen).
• Added regression tests for the discovery and request contract.
• Reviewed docs/context/architecture/catalog.md; no subsystem mechanism changed, so no fragment edit is needed.
• No secrets in the diff.

Exa's podcast documentation describes the same /search endpoint and domain filter. Pricing follows Exa's search and contents rates, with Treg reporting the actual charge. Related Treg feedback ID: 6057.