Add NVIDIA ASR NIM transcription provider

[absalonCRC]

Summary

• Add an NVIDIA ASR NIM provider wired as --api nvidia
• Use environment-driven NVIDIA_NIM_API_KEY, NVIDIA_NIM_BASE_URL, and NVIDIA_NIM_MODEL configuration
• Document the new provider in the README and add focused stdlib tests for configuration, request construction, and validation

Test plan

• PYTHONPATH=src python3 -m unittest discover -s tests
• PYTHONPATH=src python3 -m py_compile src/sapat/script.py src/sapat/transcription/nvidia.py tests/test_nvidia_transcription.py

[absalonCRC]

Status update for review:

• This companion PR supports Run NVIDIA ASR NIM with Sapat daytona/content#233 for the NVIDIA ASR NIM / Parakeet Sapat guide.
• Fresh local revalidation on 2026-05-21T11:34+02:00 passed:
  • PYTHONPATH=src python3 -m unittest discover -s tests -> 3 tests passed
  • PYTHONPATH=src python3 -m py_compile src/sapat/script.py src/sapat/transcription/nvidia.py tests/test_nvidia_transcription.py
• The current GitHub status is only waiting on external Continue agent checks; I do not see a failing test/check result.
• Maintainer edits are enabled and the implementation uses environment-based config only, with no secrets committed.

[absalonCRC]

Hi @nibzard, I tried to request review through GitHub but do not have permission to use reviewer assignment on this upstream PR. Could you route this for review when available?

This is the companion implementation for daytona/content#233. Local validation passed with unittest discovery and py_compile, maintainer edits are enabled, and the implementation keeps NVIDIA config in environment variables only.

[nibzard]

Thank you for this contribution! Your implementation was reviewed as part of a batch review of all incoming provider PRs.

We've decided to take a different approach: instead of merging individual provider PRs (which all targeted a different code architecture than what exists on main), we've refactored the entire codebase into a proper provider plugin architecture with auto-discovery and shared mixins. Your provider's API integration patterns were used as reference when building the NVIDIA NIM provider in the new architecture.

The refactor is on branch feat/provider-plugin-architecture and includes 29 providers, 176 tests, and a clean plugin system. We appreciate your contribution — it helped shape the final design. Closing this in favor of the unified approach.