feat(server): add legacy /v1/completions endpoint for the inbuilt inference server

The built-in inference server only exposed the chat protocol at
/v1/chat/completions, so OpenAI-legacy clients (text completions) could
not use it. Extract the request_config parsing, approach/model resolution,
client building and dispatch of proxy() into a shared _execute_completion()
pipeline, and add POST /v1/completions on top of it:

- prompt (str or list[str]) maps to user messages; the pipeline runs once
  per prompt and n still expands attempts into multiple choices
- returns the OpenAI legacy shape: id cmpl-*, object text_completion,
  choices[].text/logprobs/finish_reason and usage; passthrough completion
  dicts (none approach / full-response plugins / batch mode) are mapped
  choice-by-choice
- stream=true returns 400 in this first version (chat endpoint streams
  exactly as before); optillm_approach, max_tokens and other params pass
  through like on the chat endpoint
- /v1/chat/completions responses are unchanged; conversation logging,
  batch mode and error mapping (500 JSON) behave as before

Fixes #170

Author: akushonkamen

README.md

@@ -375,6 +375,21 @@ response = client.chat.completions.create(
 )
 ```
 
+The server also exposes the legacy text completions endpoint `POST /v1/completions` (returning `object: "text_completion"` responses), so OpenAI-legacy clients work against the inbuilt inference server too:
+
+```bash
+curl http://localhost:8000/v1/completions \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer $OPTILLM_API_KEY" \
+  -d '{
+    "model": "meta-llama/Llama-3.2-1B-Instruct",
+    "prompt": "Write a haiku about recursion.",
+    "max_tokens": 64
+  }'
+```
+
+Streaming (`stream: true`) is not supported on `/v1/completions` yet; use `/v1/chat/completions` for streaming responses.
+
 ### Starting the optillm proxy with an external server (e.g. llama.cpp or ollama)
 
 - Set the `OPENAI_API_KEY` env variable to a placeholder value