fix(translator): ESM-safe registry + tool-id pairing + responses max_tokens; add real-creds tests

- translator/index.js: replace require() with static side-effect imports (ESM-safe),
  lazy-init registry maps to survive circular import order
- openai-responses->openai: map max_output_tokens -> max_tokens (avoid leaking field upstream)
- gemini/antigravity -> openai: derive deterministic tool_call id from name so
  functionCall/functionResponse pair correctly (fixes provider tool-pairing 400s)
- add offline unit tests (finish-reason, usage, session-manager, ollama malformed args, const guard)
- add real-creds integration tests (provider-cases + all-formats matrix: 6 inbound formats x 4 scenarios)

Includes co-located provider registry refactor (pricing/capabilities/media providers) and sessionManager updates.

Co-authored-by: Cursor <cursoragent@cursor.com>

open-sse/translator/request/gemini-to-openai.js

@@ -97,8 +97,10 @@ function convertGeminiContent(content) {
     }
 
     if (part.functionCall) {
+      // Gemini lacks a native call id; derive a deterministic one from the name so the
+      // matching functionResponse maps to the same tool_call_id (providers require pairing).
       toolCalls.push({
-        id: `call_${Date.now()}_${Math.random().toString(36).slice(2, 8)}`,
+        id: part.functionCall.id || `call_${part.functionCall.name}`,
         type: OPENAI_BLOCK.FUNCTION,
         function: {
           name: part.functionCall.name,
@@ -110,7 +112,7 @@ function convertGeminiContent(content) {
     if (part.functionResponse) {
       return {
         role: ROLE.TOOL,
-        tool_call_id: part.functionResponse.id || part.functionResponse.name,
+        tool_call_id: part.functionResponse.id || `call_${part.functionResponse.name}`,
         content: JSON.stringify(part.functionResponse.response?.result || part.functionResponse.response || {})
       };
     }