[1.4] Backport helper service detection retries

[joaoantoniocardoso]

Summary

• Backport of helper: allow 3 attempts when detecting services #3454 (helper: allow 3 attempts when detecting services) onto 1.4-dev
• Includes the follow-up typing fixes from master (max_attempts_reached bool cast + attempts_left: Dict[int, int])
• Fixes [1.4] Helper: Available Services can permanently omit healthy BlueOS system services after boot-time probe failure #4132: Available Services could permanently omit healthy BlueOS system services after a failed first probe at boot

Test plan

• Boot BlueOS 1.4 and force early helper scans before Cable Guy / Commander / Kraken / Beacon / Ping are ready
• Confirm a failed first detect (Invalid: None - None) no longer permanently hides the service
• Confirm the service appears in /helper/latest/web_services once healthy (without restarting helper)
• Confirm timeout-heavy non-HTTP listeners still stop being retried after 3 timeout attempts

[github-actions]

Automated PR Review

0. Summary

• Verdict: LGTM

Backport for the 1.4 branch: in core/services/helper/main.py, periodic() now filters invalid entries out of Helper.KNOWN_SERVICES every 60 s so a system service that failed initial detection (e.g. wasn't ready yet at boot and burned through MAX_ATTEMPTS_LEFT) gets a fresh chance instead of being permanently skipped via the service.max_attempts_reached() short-circuit in scan_ports() (core/services/helper/main.py:410). The KEEP_BLUEOS_SERVICES_ALIVE=False branch in scan_ports() only evicts services whose TCP port stopped listening, which didn't cover this "listening but not responding correctly" case. Minimal, well-scoped fix with a comment that explains the "why".

No further comments, nice job 👍

Generated by PR Review Bot. This is advisory, a human reviewer must still approve.