M3: OpenAI chat and responses codecs, provider, registry wiring

Adds both OpenAI codecs — chat completions and the newer responses API — plus the OpenAiProvider and registry wiring, so the same session can run against OpenAI in addition to Anthropic.

crates/harness-app/src/lib.rs

@@ -19,7 +19,7 @@ use harness_core::tool::ToolRegistry;
 use harness_core::types::{
     Message, MessageId, ModelRef, Part, PartBody, PartId, Session, SessionId,
 };
-use harness_providers::{AnthropicProvider, ProviderRegistry};
+use harness_providers::{AnthropicProvider, OpenAiProvider, ProviderRegistry};
 use tokio::task::JoinHandle;
 use tokio_util::sync::CancellationToken;
 
@@ -122,6 +122,21 @@ impl EngineHandle {
         {
             providers.register(Arc::new(AnthropicProvider::new(key)));
         }
+        if let Some(key) = config
+            .providers
+            .get("openai")
+            .and_then(|p| p.api_key.clone())
+        {
+            let provider = match config
+                .providers
+                .get("openai")
+                .and_then(|p| p.base_url.clone())
+            {
+                Some(base_url) => OpenAiProvider::with_base_url(key, base_url),
+                None => OpenAiProvider::new(key),
+            };
+            providers.register(Arc::new(provider));
+        }
 
         let data_dir = dirs::data_dir()
             .unwrap_or_else(std::env::temp_dir)