docs(administration): add node maintenance & cordoning guide

[Reiers]

Context

Replaces the code approach in #1026 with operator documentation, per team consensus that cordon behavior should remain simple and user-controlled.

What's added

• New page: administration/node-maintenance.md — covers what cordoning does, how it affects location-bound pipeline tasks, and recommended workflows for:
  • Quick maintenance (restart/upgrade)
  • Planned extended maintenance
  • Node decommissioning / long outage (attach storage to another node)
• Cross-reference from troubleshooting/common-errors.md — so operators who find stuck sectors after cordoning get pointed to the guide.
• SUMMARY.md + Administration README updated with the new page.

Why docs instead of code

The discussion on #1025 and #1026 settled on keeping cordon simple by design:

• Cordon = planned maintenance, quick turnaround
• Adding drain exceptions leads to scope creep (conditional drain, timeouts, per-task policies)
• Operator architecture varies too much SP-to-SP for one-size-fits-all drain logic
• The right fix is operator awareness, not automatic behavior

This guide gives operators the information they need to plan maintenance without losing pipeline work.

Closes #1025
Refs: #1026