fix: normalize finish_reason to 'tool_calls' when tool calls are present (#379)

Some upstream providers (e.g. Antigravity) return non-standard finish_reason
values like 'other' instead of the OpenAI-standard 'tool_calls' when the
model invokes tools. This causes downstream consumers (e.g. OpenClaw) to
fail to execute tool calls, breaking agentic sub-agent workflows.

Changes:
- nonStreamingHandler: post-translation guard that normalizes finish_reason
  to 'tool_calls' when message.tool_calls is present
- sseToJsonHandler: accumulate tool_calls from streaming deltas in
  parseSSEToOpenAIResponse; extract function_call items from Responses API
  output in handleForcedSSEToJson
- openai-responses translator: use toolCallIndex to choose between
  'tool_calls' and 'stop' in flush and response.completed events

Tested: 7 scenarios (non-stream text, single/multiple tool calls, stream
text/tool calls, multi-turn tool conversation, tools present but unused)

open-sse/translator/response/openai-responses.js

@@ -364,6 +364,7 @@ export function openaiResponsesToOpenAIResponse(chunk, state) {
     // Flush: send final chunk with finish_reason
     if (!state.finishReasonSent && state.started) {
       state.finishReasonSent = true;
+      const hasToolCalls = state.toolCallIndex > 0;
       return {
         id: state.chatId || `chatcmpl-${Date.now()}`,
         object: "chat.completion.chunk",
@@ -372,7 +373,7 @@ export function openaiResponsesToOpenAIResponse(chunk, state) {
         choices: [{
           index: 0,
           delta: {},
-          finish_reason: "stop"
+          finish_reason: hasToolCalls ? "tool_calls" : "stop"
         }]
       };
     }
@@ -505,7 +506,9 @@ export function openaiResponsesToOpenAIResponse(chunk, state) {
     
     if (!state.finishReasonSent) {
       state.finishReasonSent = true;
-      state.finishReason = "stop"; // Mark for usage injection in stream.js
+      const hasToolCalls = state.toolCallIndex > 0;
+      const resolvedFinishReason = hasToolCalls ? "tool_calls" : "stop";
+      state.finishReason = resolvedFinishReason; // Mark for usage injection in stream.js
       
       const finalChunk = {
         id: state.chatId,
@@ -515,7 +518,7 @@ export function openaiResponsesToOpenAIResponse(chunk, state) {
         choices: [{
           index: 0,
           delta: {},
-          finish_reason: "stop"
+          finish_reason: resolvedFinishReason
         }]
       };