v0.7.1: Zero-touch k8s attribution + multi-harness app-context

Follow-on to v0.7.0's Portable Producer Profiles — no core/schema change.

Added

• sandbox capture — one-shot k8s pod attribution (the ~12-step manual node-side flow → one command); --container pins a named container in a multi-container pod; the sensor's libssl/libc uprobes target the pod's own /proc/<pid>/root, so model intent + DNS are captured node-side.
• sandbox watch — auto-attribution phase-2 preview: a node-side control loop that binds every pod to a run with no per-pod command (agentprov.io/run annotation, else an auto-run per UID). Honestly a kubectl poll, not yet a client-go informer; its single per-round sensor makes model-intent coverage weaker than capture.
• Egress resolves to domains — the getaddrinfo uprobe on the pod's libc + dns_query↔connect correlation shows api.deepseek.com, not just the IP.
• Multi-harness app-context (hooks bridge --harness kimi|codex|claude) — the major coding CLIs aren't black boxes; a thin per-harness adapter normalizes each one's session transcript into the existing bridge. Kimi Code (single + multi-agent delegation, merged into one wall-clock timeline) and Codex (exec_command shell tool → Bash for command-match). All three harnesses validated end-to-end against live runs.

Fixed

• Node-side tls_read reliability — close-framed HTTP responses on a reused SSL* connection were lost on long-lived workloads; a new request now flushes the pending response. tls_read went 0 → 1:1 with tls_write.

Full notes: CHANGELOG.md.