Add SambaNova as OpenAI-compatible provider

[snova-kwasia]

Why

Add SambaNova as a supported model provider in OpenWorker, enabling users to use SambaNova's cloud API for agentic workloads.

Root cause

SambaNova was not in the provider registry despite having a fully OpenAI-compatible API.

Fix

• Added SambaNova to via the existing _compat() helper (same pattern as Together AI, Fireworks, OpenRouter, etc.), pointing to https://api.sambanova.ai/v1
• Added 6 curated SambaNova models to matrix.py: Llama 3.3 70B, DeepSeek V3.1, DeepSeek V3.2, MiniMax M2.7, Gemma 4 31B, and GPT-OSS 120B

Tests

Verified sambanova is registered and all 6 models appear in the matrix:

$ uv run python -c "from coworker.providers.registry import provider_names; print('sambanova' in provider_names())"
True
$ uv run python -c "from coworker.providers.matrix import models_for_provider; print(models_for_provider('sambanova'))"
['Meta-Llama-3.3-70B-Instruct', 'DeepSeek-V3.1', 'DeepSeek-V3.2', 'MiniMax-M2.7', 'gemma-4-31B-it', 'gpt-oss-120b']

Confidence

High — follows the exact same pattern as all other OpenAI-compatible providers (Together AI, Fireworks, OpenRouter, etc.) and uses the verified official endpoint from SambaNova's docs.

Authored by a SambaChat managed agent — draft, awaiting human review.

[snova-kwasia]

Superseded by #358 (same SambaNova OpenAI-compatible provider integration). Closing to avoid duplicate PRs on the repo.