feat: replace static hardware fields with dynamic node properties

Replaces the 4 fixed hardware columns (cpu_count, cpu_model, ram_gb,
disk_gb) with a flexible properties system. Each property has a key,
value, icon (from a curated Lucide picker), and a visibility toggle
that controls whether it appears on the canvas node card.

- Backend: add `properties` JSON column to Node model; data migration
  converts existing hardware rows to properties with correct icons
  (idempotent, old columns kept for safety)
- Backend: add `properties` to NodeBase, NodeUpdate, NodeSave schemas
  and canvasSerializer so values survive canvas save/load
- Frontend: add NodeProperty type; new propertyIcons.ts registry (20
  icons); BaseNode renders visible properties with legacy hardware
  fallback for unmigrated nodes
- Frontend: DetailPanel gains interactive properties section (add /
  edit / remove / toggle visibility / icon picker) replacing the
  read-only hardware block; hardware section removed from NodeModal
- Tests: 6 migration tests, 7 API tests, 8 DetailPanel property tests,
  6 BaseNode render/fallback tests, 9 propertyIcons util tests

backend/app/db/models.py

@@ -42,6 +42,7 @@ class Node(Base):
     ram_gb: Mapped[float | None] = mapped_column(Float, nullable=True)
     disk_gb: Mapped[float | None] = mapped_column(Float, nullable=True)
     show_hardware: Mapped[bool] = mapped_column(Boolean, default=False)
+    properties: Mapped[list[Any]] = mapped_column(JSON, default=list)
     width: Mapped[float | None] = mapped_column(Float, nullable=True)
     height: Mapped[float | None] = mapped_column(Float, nullable=True)
     bottom_handles: Mapped[int] = mapped_column(Integer, default=1)
@@ -49,7 +50,6 @@ class Node(Base):
     response_time_ms: Mapped[int | None] = mapped_column(Integer)
     created_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=_now)
     updated_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=_now, onupdate=_now)
-
     children: Mapped[list["Node"]] = relationship("Node", back_populates="parent")
     parent: Mapped["Node | None"] = relationship("Node", back_populates="children", remote_side=[id])