feat(catalog): make the four international vendors selectable when listing a key

[argszero]

Summary

The four international vendors were in the model catalogue but could not be put
on the shelf. Tracked as rant 2026-10-16T14:08:13.998860.

config/config.example.toml had [[models]] rows for openai, anthropic,
google and xai but no [[providers]] row and no [[plans]] row for any of
them. The listing form builds its vendor dropdown from the providers that have a
plan (ui/js/app.js: [...new Set(planList().map((pl) => pl.provider))]), so the
four never appeared in the vendor -> plan -> model cascade and no key for those
models could be listed at all.

Changes

• config/config.example.toml: four [[providers]] (OpenAI / Anthropic /
  Google Gemini / xAI, country = "US") and one pay-as-you-go [[plans]] per
  vendor, built from each vendor's official base URL. Endpoint paths are the ones
  the gateway derives (src/gateway.rs::resolve_endpoint): openai_chat ->
  {base}/chat/completions, anthropic -> {base}/v1/messages, responses ->
  {base}/responses. Base URLs probed 2026-10-16: https://api.openai.com/v1,
  https://api.anthropic.com, https://generativelanguage.googleapis.com/v1beta/openai,
  https://api.x.ai/v1.
• ui/js/data.js: the four plan ids mirrored into PLANS (the listing form's
  fallback table; it must stay equal to the config's plan ids) and the four
  display labels into PROVIDER_LABELS.
• src/catalog_gate.rs: PLAN_COUNT 12 -> 16. This is the consistency gate's
  positive control, the number it asks to be updated deliberately whenever the
  plan list changes.
• src/config.rs / src/main.rs: the example-config provider count assertions
  6 -> 10.
• ui/README.md: the listing form's built-in plan roster now names the four.
• ui/index.html: bumped the data.js cache-bust token.

No model rows and no prices are touched: MODEL_COUNT and MARKET_COUNT are
unchanged, and the four vendors' models (gpt-5.6-sol, claude-opus-5,
gemini-3.1-pro, grok-4.6 / grok-4.7) were already in the catalogue.

Related Issue

None — this is tracked as a rant in this deployment's queue rather than as a
GitHub issue.

Tests

• Existing tests pass — cargo test: 564 passed / 0 failed. The count is
  unchanged; the catalogue consistency gate (src/catalog_gate.rs) covers the
  plan-id parity this change exercises, and its PLAN_COUNT control was updated
  with the data.
• cargo fmt --check clean; cargo clippy --all-targets clean.