feat: add TokenLab provider

[hedging8563]

Summary

• Add tokenlab as a Continue LLM provider backed by the OpenAI-compatible chat endpoint at https://api.tokenlab.sh/v1/.
• Register the provider in the core LLM class list and add it to the shared OpenAI-compatible provider test suite.
• Add TokenLab docs with YAML/JSON examples, example model IDs, and a note that TokenLab also exposes native Responses, Anthropic Messages, Gemini generateContent, embeddings, images, videos, audio, realtime, and translation endpoints for direct API callers.

Validation

• cd core && npm run vitest -- llm/llms/OpenAI-compatible-core.vitest.ts
• cd core && npm run tsc:check
• npx prettier --check core/llm/llms/TokenLab.ts core/llm/llms/index.ts core/llm/llms/OpenAI-compatible-core.vitest.ts docs/customize/model-providers/more/tokenlab.mdx docs/customize/model-providers/overview.mdx
• node -e "fetch('https://api.tokenlab.sh/v1/models').then(r=>r.json().then(j=>{const ids=(j.data||[]).map(m=>m.id); console.log(r.status, ['gpt-5.5','claude-sonnet-5','gemini-3.5-flash','qwen3.7-max'].every(id=>ids.includes(id)), ids.length);}))"

---

Supersedes #12959. The previous PR was closed while fixing contributor commit metadata; this replacement branch preserves the same scoped diff with commits authored by hedging8563 <45883076+hedging8563@users.noreply.github.com>.

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[hedging8563]

I have read the CLA Document and I hereby sign the CLA

[hedging8563]

recheck

[hedging8563]

CI update: I pushed refresh commit 6806b2e after the CLA was completed. The Windows VSIX job now passes, and the remaining required failure is jetbrains-tests only.\n\nI checked the failure artifact: the regular JetBrains unit test report shows 53/53 passing; the single failure is the JetBrains integration UI test Autocomplete > testAutocomplete(), where the assertion text.contains("TEST_LLM_RESPONSE_0") is false. This PR only changes the TokenLab OpenAI-compatible provider registration/tests and provider docs, so it does not touch the JetBrains autocomplete integration path. Could a maintainer rerun or advise on that integration-test failure?