fix: replace netstat with /proc/net/tcp for port availability checks

[barnabasbusa]

The fluent-bit debug container (distroless:debug) does not include netstat, causing the logs collector port check to always fail on k8s. Use /proc/net/tcp which works on any Linux container without tool deps.

When it broke: Commit be932f7 ("impl k8s fluentbit collector availability check", Feb 18, 2025), merged via PR #2653 at commit 05e3813 ("feat:
fluent bit logs collector in k8s backend", Feb 22, 2025). This has been broken since day one of the k8s logs collector feature - it was never
working.

Root cause: The port availability check runs netstat -anp tcp | grep LISTEN | grep 9713 via kubectl exec inside the fluent/fluent-bit:4.0.0-debug
container. That image is based on Google's distroless:debug, which has a minimal busybox that does not include netstat. The engine and API
containers are Alpine-based and have busybox netstat, so their checks work fine.

The fix: Replaced the netstat command with a /proc/net/tcp check in both the Kubernetes and Docker backends:
grep -i ':25F1 ' /proc/net/tcp /proc/net/tcp6 2>/dev/null | grep -qi ' 0A '

• Reads the Linux proc filesystem directly (always available, zero tool dependency)
• 25F1 is the hex representation of the port number
• 0A is the TCP LISTEN state in /proc/net/tcp

[barnabasbusa]

Closing as it is now part of #2891