feat: add Venice provider proxy support

[toshipepe]

Summary

• add a fixed /venice localhost route that forwards Responses and Chat Completions calls to https://api.venice.ai/api/v1
• record Venice separately while reusing OpenAI-compatible token, cache, model, and streaming-final usage parsing
• keep provider request IDs local for future reconciliation and prove they are excluded from optional cloud sync
• document environment-only VENICE_API_KEY setup in user-level Codex config, with no prompt or response persistence
• leave unknown Venice models outside authoritative priced totals

Limitations

• synthetic/test-covered only; no paid Venice request or invoice reconciliation was performed
• provider-catalog pricing and billing reconciliation remain separate follow-up phases
• no API keys, prompts, responses, private cloud code, or local roadmap notes are included

Sources

• Venice API reference: https://docs.venice.ai/api-reference/api-spec
• Venice Codex guide: https://docs.venice.ai/guides/integrations/codex-cli
• Codex advanced configuration: https://learn.chatgpt.com/docs/config-file/config-advanced

Verification

• node scripts/privacy-check.mjs --ci
• python3 tests/test_basic.py (29 passed)
• cd ts && npm test (114 passed)
• node scripts/privacy-check.mjs --precommit
• exact staged diff reviewed; ignored .codex/NEXT_PRODUCT_CHECKLIST.md remained uncommitted