[codex] Support multi-pod application deployment

[hhhhsc701]

What changed

• Add Redis-backed distributed agent run state, cancellation, heartbeat, and stream replay support.
• Make shared file and skill writes safe across replicas.
• Add process-local health endpoints and multi-replica Kubernetes configuration, PDBs, and RWX storage validation.
• Preserve frontend run errors and make shared locale configuration writes atomic.

Why

The application services need to run with multiple replicas without losing run state, corrupting shared files, or routing resume/stop requests to the wrong pod.

Impact

Runtime state now depends on Redis in multi-pod deployments. Application charts default to three replicas and require shared RWX storage.

Validation

• 127 targeted backend/SDK tests passed.
• 8 frontend Node tests passed.
• git diff --check passed.

Stack

Base PR in the three-PR split. The Helm release split is stacked on this branch.

[codecov]

Codecov Report

❌ Patch coverage is 83.23587% with 86 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
backend/services/agent_service.py	63.04%	47 Missing and 21 partials ⚠️
backend/services/streaming_channel.py	74.19%	4 Missing and 4 partials ⚠️
backend/services/northbound_service.py	87.50%	3 Missing and 1 partial ⚠️
sdk/nexent/utils/atomic_file.py	90.00%	3 Missing and 1 partial ⚠️
backend/northbound_service.py	0.00%	1 Missing ⚠️
backend/services/runtime_state_service.py	99.11%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!