Add Ingero - eBPF GPU causal observability for production ML

[dml37]

Adds Ingero to Evaluation and Monitoring.

Ingero is an open-source (Apache 2.0 + GPL-2.0) eBPF agent and MCP server purpose-built for production ML observability. It traces the causal chain from Linux kernel events through CUDA API calls to Python source lines, so a stall in forward() can be attributed to (e.g.) cudaMalloc spiking under host-side CPU contention.

• One binary, <2% overhead, zero code changes, runs as a DaemonSet on Kubernetes.
• Used to debug PyTorch / vLLM training and inference stalls.
• Already featured in awesome-ebpf, awesome-observability, awesome-opentelemetry, awesome-sre-tools, Awesome-GPU, awesome-mcp-servers, and others.

Repo: https://github.com/ingero-io/ingero