fix: do not fatally abort install on nm-online connectivity check

[rustydb]

Problem:

The management-network readiness check ran nm-online -x, which exits immediately instead of waiting, and treated any non-zero result as fatal. On the already-installed first-boot path this check runs before the RancherD config is written, so a management link that is briefly offline while the bond re-converges aborts the whole install and leaves the node with no role until manual repair.

Solution:

Two changes:

• Wait for the network instead of sampling once: use nm-online -s -t 45 instead of -x. -s waits for NetworkManager startup to complete (interfaces finished activating) rather than for global connectivity, which an isolated management network may never reach. The 45s timeout gives a slow-rate LACP bond headroom: lacp_rate=slow emits an LACPDU only every 30s, so a single missed interval can cost ~30s before the bond syncs. Because -s returns as soon as startup completes, this budget only costs time on a genuinely slow or degraded link.
• Make the check non-fatal: log a warning and continue instead of returning an error, so a slow or briefly-degraded management link no longer aborts the install (and, on the installed-node path, no longer strands the node without a RancherD config).

This is observed on both LACP/802.3ad and single-NIC active-backup bonds, and across install methods, so it is not specific to any one bond mode.

Related Issue(s):

Refs: harvester/harvester#10885

Test plan:

Additional documentation or context