Add semantic source-of-truth tagging + grouped layer toggles

[marota]

Summary

The overflow graph and its interactive HTML viewer now expose
semantic layer toggles driven by source-of-truth boolean flags
the recommender propagates onto graph nodes and edges, replacing
brittle colour / shape heuristics.

OverflowGraph (alphaDeesp/core/graphs/overflow_graph.py)

• set_hubs_shape tags every hub with is_hub=True (and also
  in_red_loop=True + on_constrained_path=True so hubs always
  show under those layers regardless of how the recommender's
  lists are built).
• New tag_red_loops(lines, nodes) and
  tag_constrained_path(lines, nodes) helpers consume the
  recommender's pre-computed lists.
• highlight_significant_line_loading tags is_monitored=True
  on every monitored entry and is_overload=True on the strict
  subset that is contingency-overloaded — is_overload is a
  strict subset of is_monitored.

Interactive HTML viewer (alphaDeesp/core/interactive_html.py)

• _build_layer_index scans data-attr-* flags on both nodes
  and edges and emits semantic layers (Hubs, Red-loop paths,
  Constrained path, Overloads, Low margin lines).
• Layers are grouped into three labelled sections — Structural
  Paths / Individual entities properties / Flow redispatch
  values — via _LAYER_SECTIONS and rendered in _SECTION_ORDER
  with <h3 class=\"layer-section-header\"> markers.
• Membership-based dim model: an element is visible iff at
  least one of its claiming layers is checked; opacity drops to
  0.12 otherwise. No display:none — click / hover stay
  live so the operator can still hit dimmed elements.
• Endpoint-node inclusion: edge-only layers also dim /
  highlight the bbox-endpoint nodes of their claimed edges, so
  toggling "Overloads" alone keeps the substations the
  overload connects visible.
• Select-all / Unselect-all link row above the layer list,
  dispatching the same change event as individual checkboxes.
• _align_edge_ids_with_svg() regression guard: graphviz's
  SVG and JSON outputs assign edge ids in independent orders, so
  data-source / data-target on <g class=\"edge\"> ended
  up disagreeing with the edge's <title> endpoints. We now
  re-key the JSON edge model from the SVG title strings.

Test plan

• pytest alphaDeesp/tests/test_overflow_graph.py — new
  assertions for the is_hub / in_red_loop / is_monitored
  / is_overload / on_constrained_path attributes, the
  strict overload-subset semantics, and the new taggers.
• pytest alphaDeesp/tests/test_interactive_html.py — new
  semantic layer keys, section grouping, membership dim model,
  endpoint-node inclusion, edge-id alignment guard, and the
  Select-all / Unselect-all row.
• Downstream verification: Expert_op4grid_recommender
  (companion PR) wires the lists into the new taggers.
• Downstream verification: Co-Study4Grid (companion PR)
  injects the action-pin overlay on top of the viewer.

🤖 Generated with Claude Code