featu: support passing "extra body" throught to providers

# What does this PR do?
Allows passing through extra_body parameters to inference providers.


closes #2720

## Test Plan
CI and added new test

docs/static/deprecated-llama-stack-spec.yaml

@@ -1098,7 +1098,7 @@ paths:
         content:
           application/json:
             schema:
-              $ref: '#/components/schemas/OpenAIChatCompletionRequest'
+              $ref: '#/components/schemas/OpenAIChatCompletionRequestWithExtraBody'
         required: true
       deprecated: true
   /v1/openai/v1/chat/completions/{completion_id}:
@@ -1167,7 +1167,7 @@ paths:
         content:
           application/json:
             schema:
-              $ref: '#/components/schemas/OpenAICompletionRequest'
+              $ref: '#/components/schemas/OpenAICompletionRequestWithExtraBody'
         required: true
       deprecated: true
   /v1/openai/v1/embeddings:
@@ -5575,7 +5575,7 @@ components:
       title: OpenAIResponseFormatText
       description: >-
         Text response format for OpenAI-compatible chat completion requests.
-    OpenAIChatCompletionRequest:
+    OpenAIChatCompletionRequestWithExtraBody:
       type: object
       properties:
         model:
@@ -5717,7 +5717,7 @@ components:
       required:
         - model
         - messages
-      title: OpenAIChatCompletionRequest
+      title: OpenAIChatCompletionRequestWithExtraBody
       description: >-
         Request parameters for OpenAI-compatible chat completion endpoint.
     OpenAIChatCompletion:
@@ -5885,7 +5885,7 @@ components:
         - model
         - input_messages
       title: OpenAICompletionWithInputMessages
-    OpenAICompletionRequest:
+    OpenAICompletionRequestWithExtraBody:
       type: object
       properties:
         model:
@@ -5973,18 +5973,6 @@ components:
         user:
           type: string
           description: (Optional) The user to use.
-        guided_choice:
-          type: array
-          items:
-            type: string
-          description: >-
-            (Optional) vLLM-specific parameter for guided generation with a list of
-            choices.
-        prompt_logprobs:
-          type: integer
-          description: >-
-            (Optional) vLLM-specific parameter for number of log probabilities to
-            return for prompt tokens.
         suffix:
           type: string
           description: >-
@@ -5993,7 +5981,7 @@ components:
       required:
         - model
         - prompt
-      title: OpenAICompletionRequest
+      title: OpenAICompletionRequestWithExtraBody
       description: >-
         Request parameters for OpenAI-compatible completion endpoint.
     OpenAICompletion: