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-runtime/src/tool/builtin/execute_skill.rs

@@ -4,7 +4,7 @@ use async_trait::async_trait;
 use serde_json::{json, Value};
 use zclaw_types::{Result, ZclawError};
 
-use crate::tool::{Tool, ToolContext};
+use crate::tool::{Tool, ToolContext, ToolConcurrency};
 
 pub struct ExecuteSkillTool;
 
@@ -42,6 +42,10 @@ impl Tool for ExecuteSkillTool {
         })
     }
 
+    fn concurrency(&self) -> ToolConcurrency {
+        ToolConcurrency::Exclusive
+    }
+
     async fn execute(&self, input: Value, context: &ToolContext) -> Result<Value> {
         let skill_id = input["skill_id"].as_str()
             .ok_or_else(|| ZclawError::InvalidInput("Missing 'skill_id' parameter".into()))?;