perf(runtime): Hermes Phase 1-3 — prompt caching + parallel tools + smart retry

Phase 1: Anthropic prompt caching
- Add cache_control ephemeral on system prompt blocks
- Track cache_creation/cache_read tokens in CompletionResponse + StreamChunk

Phase 2A: Parallel tool execution
- Add ToolConcurrency enum (ReadOnly/Exclusive/Interactive)
- JoinSet + Semaphore(3) for bounded parallel tool calls
- 7 tools annotated with correct concurrency level
- AtomicU32 for lock-free failure tracking in ToolErrorMiddleware

Phase 2B: Tool output pruning
- prune_tool_outputs() trims old ToolResult > 2000 chars to 500 chars
- Integrated into CompactionMiddleware before token estimation

Phase 3: Error classification + smart retry
- LlmErrorKind + ClassifiedLlmError for structured error mapping
- RetryDriver decorator with jittered exponential backoff
- Kernel wraps all LLM calls with RetryDriver
- CONTEXT_OVERFLOW recovery triggers emergency compaction in loop_runner

crates/zclaw-types/src/error.rs

@@ -223,6 +223,33 @@ impl Serialize for ZclawError {
 /// Result type alias for ZCLAW operations
 pub type Result<T> = std::result::Result<T, ZclawError>;
 
+/// LLM 调用错误的细粒度分类，指导重试和恢复策略
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum LlmErrorKind {
+    Auth,
+    AuthPermanent,
+    BillingExhausted,
+    RateLimited,
+    Overloaded,
+    ServerError,
+    Timeout,
+    ContextOverflow,
+    ModelNotFound,
+    Unknown,
+}
+
+/// 分类后的 LLM 错误，附带恢复提示
+#[derive(Debug, Clone)]
+pub struct ClassifiedLlmError {
+    pub kind: LlmErrorKind,
+    pub retryable: bool,
+    pub should_compress: bool,
+    pub should_rotate_credential: bool,
+    pub retry_after: Option<std::time::Duration>,
+    pub message: String,
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;