Add workload failure reason for terminal agent runtime failure

[rowan-stein]

User request

Support prompt degraded-dialog handling for terminal agent-processing failures.

Umbrella issue: agynio/agents-orchestrator#221

Specification

Add a structured workload failure reason to the Runners API so downstream services can distinguish terminal agent-processing/runtime failures from ordinary transient start/runtime failures.

Suggested enum addition in proto/agynio/api/runners/v1/runners.proto:

WORKLOAD_FAILURE_REASON_AGENT_RUNTIME_FAILURE = 6;

Semantics:

• The agent process reached a terminal, non-retryable processing error while handling a thread message.
• This is distinct from start failure, crashloop, image pull failure, config invalid, and generic runtime lost.
• Agents Orchestrator may use this reason to degrade the owning thread promptly.

Acceptance criteria

• WorkloadFailureReason includes a new value for terminal agent runtime/processing failure.
• Generated bindings are updated if this repo stores generated artifacts.
• Backward compatibility is preserved; existing enum values are not renumbered.