perf: runtime lookup and polling optimizations

[ibrahimkizilarslan]

1. Docker Container ID Cache & Filters (P1): Added an LRU-style cache with a 30s TTL for container ID lookups. Changed getContainerID to use Docker's filter API instead of listing all containers, preventing O(N) API calls per operation.
2. Kubernetes Pod Lookup Optimization (P2): Modified findPod to attempt exact name matching (O(1)) and label-based selection (server-side filtering) before falling back to the expensive namespace-wide pod list.
3. Docker Exec Polling Backoff (P3): Replaced the fixed 100ms polling loop in ExecCommand with an exponential backoff starting at 50ms up to 500ms to reduce Docker daemon load.
4. Kubernetes Ephemeral Wait Backoff (P4): Introduced exponential backoff for ephemeral container readiness checks, reducing kube-apiserver load.
5. History Memory Leak Fix (P5): Fixed unbounded array growth in the chaos engine event history by allocating a new slice when trimming, allowing the old array to be GC'd.