fix(kiro): improve direct session cache reuse

Reshape Kiro direct requests so resumed client sessions reuse Kiro's
cache-affinity fields instead of starting unrelated CodeWhisperer
conversations.

- keep conversationState.conversationId stable when the client sends an
  explicit session id (x-session-id, session_id, conversation_id, Claude
  Code session metadata)
- add a stable conversationState.agentContinuationId per Kiro session
- send conversationState.agentTaskType: "vibe" and agentMode: "vibe",
  matching the normal Kiro CLI/KAS chat path
- move Kiro thinking instructions into Kiro-compatible systemPrompt /
  additionalModelRequestFields instead of generic top-level thinking
- keep volatile timestamp context out of the top-level systemPrompt; it
  remains only in user content fallback
- suppress additionalModelRequestFields for legacy 4.5-era Claude/Kiro
  models that reject it, while defaulting future Claude/Kiro model ids
  to supported
- preserve Kiro meteringEvent credit usage internally for accounting
  without leaking provider-specific fields into OpenAI-compatible usage
- prevent unrelated headerless Kiro requests from sharing one
  connection-wide continuation
- cap/evict continuation sessions so long-running processes do not grow
  the continuation map unbounded
- treat generated headerless Kiro sessions as one-shot so they do not
  evict real explicit-session continuations
- keep credit-only Kiro metering valid for internal persistence when
  token metrics are unavailable

open-sse/translator/index.js

@@ -103,8 +103,16 @@ export function translateRequest(sourceFormat, targetFormat, model, body, stream
     }
   }
 
-  // Normalize thinking to the target provider-native format (config-driven, capability-aware)
-  applyThinking(targetFormat, model, result, provider, thinkingIntent);
+  // Normalize thinking to the target provider-native format (config-driven, capability-aware).
+  // Kiro's GenerateAssistantResponse request does not accept the generic top-level
+  // `thinking` field; its translators map thinking intent to KAS-compatible
+  // systemPrompt/additionalModelRequestFields instead.
+  const kiroThinkingMappedByTranslator =
+    targetFormat === FORMATS.KIRO &&
+    (sourceFormat === FORMATS.OPENAI || sourceFormat === FORMATS.CLAUDE);
+  if (!kiroThinkingMappedByTranslator) {
+    applyThinking(targetFormat, model, result, provider, thinkingIntent);
+  }
 
   // Always normalize to clean OpenAI format when target is OpenAI
   // This handles hybrid requests (e.g., OpenAI messages + Claude tools)