feat: Add DeepInfra as an inference provider

[Thachnh]

What’s in this PR

This PR adds DeepInfra as a new inference provider for conversational and text-generation tasks. The integration is via DeepInfra’s OpenAI-compatible API endpoints.

Test

pnpm --filter @huggingface/inference test "test/InferenceClient.spec.ts" -t "DeepInfra"

> @huggingface/[email protected] test /home/thach/huggingface/huggingface.js/packages/inference
> vitest run --config vitest.config.mts test/InferenceClient.spec.ts -t DeepInfra


 RUN  v0.34.6 /home/thach/huggingface/huggingface.js/packages/inference

 ✓ test/InferenceClient.spec.ts (128) 1042ms
   ✓ InferenceClient (128) 1042ms
     ↓ backward compatibility (1) [skipped]
       ↓ works with old HfInference name [skipped]
     ...
     ✓ DeepInfra (3) 1041ms
       ✓ chatCompletion 988ms
       ✓ chatCompletion stream
       ✓ textGeneration 1002ms
     ...

 Test Files  1 passed (1)
      Tests  3 passed | 125 skipped (128)
   Start at  22:16:13
   Duration  3.34s (transform 1.09s, setup 23ms, collect 975ms, tests 1.04s, environment 0ms, prepare 771ms)

What’s next

Follow-up PRs will expand DeepInfra support to additional tasks such as text-to-image and text-to-audio.