JobObject HANDLE leaks on server crash-restart

[kKaskak]

server.rs:48 creates job_main_process but never stores it on self or calls CloseHandle. On crash → restart via server.rs:235, another job object is created. Each crash leaks a kernel HANDLE. On older Windows (single-job), every assign after the first fails silently.

[kKaskak]

Verdict: valid.

Each start() creates a raw JobObject handle and neither stores it nor closes it. Normal startup leaks one handle; each crash restart leaks another. Reassigning the already-running parent into new jobs is also fragile on older/hosted Windows job configurations. Store/close the handle or move ownership into a small RAII wrapper.