feat: upgrade rig-core to 0.28 and fix OpenAI Responses API multi-turn

Breaking Changes:
- Upgrade rig-core from 0.27 to 0.28
- Fix ToolCallDeltaContent API change in bedrock streaming

OpenAI Responses API Fix:
- Add simplify_history_for_openai_reasoning() to handle strict pairing requirements
- Add ensure_history_starts_with_user() for history truncation edge cases
- Strip reasoning/tool_calls/tool_results from history, keep only User/Assistant text
- Prevents "rs_... without required following item" and "fc_... without required reasoning item" errors

New Tools:
- K8sOptimizeTool: Static analysis of K8s resource requests/limits
- K8sCostsTool: Cost estimation for K8s workloads
- K8sDriftTool: Detect drift between manifests and cluster state
- PrometheusDiscoverTool: Auto-discover Prometheus in K8s clusters
- PrometheusConnectTool: Connect to Prometheus for live metrics
- RetrieveOutputTool/ListOutputsTool: RAG retrieval for compressed tool outputs

New Analyzer Module:
- src/analyzer/k8s_optimize/: Complete K8s resource optimization framework
- Static analysis rules (K8S_OPT_001 through K8S_OPT_010)
- Prometheus metrics client for live analysis
- Cost calculation and trend analysis
- Terraform kubernetes resource parsing

Dependency Updates:
- yaml-rust2: 0.9 → 0.11
- inquire: 0.7 → 0.9
- rustsec: 0.30 → 0.31
- reqwest: 0.12 → 0.13
- rand: 0.8 → 0.9
- schemars: 0.8 → 1
- termimad: 0.30 → 0.34
- New: kube 2.0, k8s-openapi 0.26, hcl-rs 0.19.4

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: 2 people

.cargo/config.toml

@@ -1,6 +1,9 @@
 [build]
 rustflags = ["-C", "target-cpu=native"]
 
+# On macOS, the default linker (ld64) is already quite fast
+# For Linux, you could use mold: rustflags = ["-C", "link-arg=-fuse-ld=mold"]
+
 [alias]
 b = "build"
 br = "build --release"