OpenAI cleanups (#32597)

Release Notes:

- openai: Remove support for deprecated o1-preview and o1-mini models 
- openai: Support streaming for o1 model

crates/language_models/src/provider/open_ai.rs

@@ -438,7 +438,7 @@ pub fn into_open_ai(
         stream,
         stop: request.stop,
         temperature: request.temperature.unwrap_or(1.0),
-        max_tokens: max_output_tokens,
+        max_completion_tokens: max_output_tokens,
         parallel_tool_calls: if model.supports_parallel_tool_calls() && !request.tools.is_empty() {
             // Disable parallel tool calls, as the Agent currently expects a maximum of one per turn.
             Some(false)
@@ -648,8 +648,6 @@ pub fn count_open_ai_tokens(
             | Model::FourPointOneMini
             | Model::FourPointOneNano
             | Model::O1
-            | Model::O1Preview
-            | Model::O1Mini
             | Model::O3
             | Model::O3Mini
             | Model::O4Mini => tiktoken_rs::num_tokens_from_messages(model.id(), &messages),