Helm chart: give bare-metal Kura fleet MHCs the full OS-reinstall window

Bare-metal Kura fleet MachineHealthChecks in the chart now use nodeStartupTimeout: 90m (was 30m) for Dedibox, OVHcloud, and Scaleway Elastic Metal fleets. The previous 30-minute window was shorter than the 45–60 minute OS reinstall those providers take, causing the MHC to repeatedly mark newly provisioned boxes unhealthy and trigger wipe-and-reinstall loops, which could wedge helm upgrade --wait rollouts. The longer startup window clears the reinstall plus boot with margin, while still leaving the 10-minute unhealthyConditions remediation active for nodes that fail after joining.