Clean up: remove remaining multi-orchestrator scaffolding

[hw-native-sys-bot]

Context

PR #437 hardcoded orchestrators[0] and removed the thread_local pto2_current_orch_idx / pto2_set_orch_thread_idx mechanism. However, several pieces of multi-orchestrator scaffolding remain in the codebase:

Remaining code to remove

• PTO2_MAX_ORCH_THREADS constant (currently 4) in all pto_runtime2.h headers
• PTO2Runtime::orchestrators[] array — can become a single PTO2OrchestratorState orchestrators[1] (or a plain member)
• PTO2Runtime::orch_count field and its validation logic in pto2_runtime_create_from_sm
• orch_idx parameter passed to orch_func_(args, orch_thread_num_, orch_idx) in executor — always 0 now
• orch_thread_num_ / sched_thread_num_ split logic in aicpu_executor.cpp (orchestrator vs scheduler thread roles)
• Multi-orchestrator docs references in SUBMIT_BY_CLUSTER.md mentioning pto2_current_orch_idx
• perf_aicpu_set_orch_thread_idx (static __thread) in performance_collector_aicpu.cpp — same pattern, also uses thread-local

Files

• src/a2a3/runtime/aicpu_build_graph/runtime/pto_runtime2.h
• src/a2a3/runtime/tensormap_and_ringbuffer/runtime/pto_runtime2.h
• src/a5/runtime/tensormap_and_ringbuffer/runtime/pto_runtime2.h
• src/a2a3/runtime/aicpu_build_graph/aicpu/aicpu_executor.cpp
• src/a2a3/runtime/tensormap_and_ringbuffer/aicpu/aicpu_executor.cpp
• src/a5/runtime/tensormap_and_ringbuffer/aicpu/aicpu_executor.cpp
• src/{a2a3,a5}/platform/src/aicpu/performance_collector_aicpu.cpp
• src/{a2a3,a5}/runtime/tensormap_and_ringbuffer/docs/SUBMIT_BY_CLUSTER.md