fix: add missing nvidia case in explicit provider switch

[elseasama]

Summary

• The nvidia provider is defined in ProvidersConfig and handled in the model-name fallback detection, but was missing from the explicit provider switch in CreateProvider()
• Setting PICOCLAW_AGENTS_DEFAULTS_PROVIDER=nvidia silently falls through to the fallback, which only matches if the model name contains "nvidia" — models like meta/llama-3.1-405b-instruct on NVIDIA's API are never matched
• Adds case "nvidia": to resolve API key, base URL, and proxy from cfg.Providers.Nvidia, consistent with all other providers

Test plan

• Set provider: "nvidia" and a non-nvidia model name (e.g. meta/llama-3.1-405b-instruct)
• Verify the agent connects and responds successfully
• Verify existing model-name fallback still works when no explicit provider is set