feat(providers): Groq now uses LiteLLM openai-compat (#1303)

Groq has never supported raw completions anyhow. So this makes it easier
to switch it to LiteLLM. All our test suite passes.

I also updated all the openai-compat providers so they work with api
keys passed from headers. `provider_data`

## Test Plan

```bash
LLAMA_STACK_CONFIG=groq \
   pytest -s -v tests/client-sdk/inference/test_text_inference.py \
   --inference-model=groq/llama-3.3-70b-versatile --vision-inference-model=""
```

Also tested (openai, anthropic, gemini) providers. No regressions.

llama_stack/templates/dev/run.yaml

@@ -29,6 +29,11 @@ providers:
     provider_type: remote::gemini
     config:
       api_key: ${env.GEMINI_API_KEY:}
+  - provider_id: groq
+    provider_type: remote::groq
+    config:
+      url: https://api.groq.com
+      api_key: ${env.GROQ_API_KEY:}
   - provider_id: sentence-transformers
     provider_type: inline::sentence-transformers
     config: {}
@@ -241,6 +246,31 @@ models:
   provider_id: gemini
   provider_model_id: gemini/text-embedding-004
   model_type: embedding
+- metadata: {}
+  model_id: meta-llama/Llama-3.1-8B-Instruct
+  provider_id: groq
+  provider_model_id: groq/llama3-8b-8192
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.1-8B-Instruct
+  provider_id: groq
+  provider_model_id: groq/llama-3.1-8b-instant
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3-70B-Instruct
+  provider_id: groq
+  provider_model_id: groq/llama3-70b-8192
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.3-70B-Instruct
+  provider_id: groq
+  provider_model_id: groq/llama-3.3-70b-versatile
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.2-3B-Instruct
+  provider_id: groq
+  provider_model_id: groq/llama-3.2-3b-preview
+  model_type: llm
 - metadata:
     embedding_dimension: 384
   model_id: all-MiniLM-L6-v2