Fix use-after-scope in RandomX job thread task lambda

[MoneroOcean]

Motivation

• Fix a high-severity use-after-scope bug where a stack-local std::atomic<int> unique_counter was captured by reference in thread-pool tasks and could be accessed after set_job() returned.
• The dangling counter produced stale thread_id values that were then used to index new_inputs and m_vm, risking crashes or native memory corruption.

Description

• Remove the local unique_counter from Core::set_job() to eliminate the dangling reference risk.
• Change the worker lambda to accept the int thread_id parameter supplied by ctpl::thread_pool and stop capturing the counter.
• Preserve all existing hashing logic and control flow so runtime behavior remains the same while removing the unsafe pattern.

Testing

• Ran automated source searches with rg to confirm unique_counter is no longer referenced and rx_thread_id usage remains consistent, and the searches succeeded.
• Inspected the modified region with sed -n '300,350p' moner-job.cpp and nl -ba moner-job.cpp | sed -n '330,360p' to verify the lambda signature and captures were updated as intended.
• Verified the diff shows the counter removal and lambda change; these automated checks completed successfully.

---

Codex Task