fix(reconciler): nodeHasCapacity uses wrong variable semantics

[Anhydrite]

Problem

In the reconciler's nodeHasCapacity function, the variable named freeCPU actually stores used_cpu from the database, which inverts the comparison logic.

Current Code

File: /root/AnhyWhere/internal/cluster/reconciler.go (lines 385-414)

The Bug

The comparison logic is inverted:

• If usedCPU = 8 and needCPU = 2, then freeCPU (8) < needCPU (2) is false, so it returns true (allowing dispatch)
• But the actual available CPU should be totalCPU - usedCPU, not just usedCPU

This means:

• Nodes with HIGH usage are incorrectly allowed to receive new VMs
• Nodes with LOW usage are incorrectly blocked

Expected Behavior

Impact

VMs may be dispatched to already-full nodes, causing:

• VM startup failures
• Resource exhaustion
• Potential crashes on the target node

Priority

Medium — This is a correctness bug but the optimistic fallback (return true on error) partially masks it in practice.