fix(topology): count rigid bodies in compute_mobility, expand hyperedges as cliques

[HugoFara]

Resolves the hyperedge convention mismatch the 1.2.0 graph tutorial documented instead of fixing.

What was wrong

compute_mobility / compute_dof counted 1 + len(edges) + len(hyperedges) links and len(nodes) joints. That is only true of the topology catalog's hyperedge-only graphs. For any geometric graph — a coupler point, a Hyperedge labelling a triangle of edges, Linkage.to_hypergraph() output for a FixedDyad, every classical walker leggedsnake ships (Walker.dof wraps this) — the answer was nonsense:

graph	before	after
tutorial coupler four-bar (hyperedge + 3 inner edges)	8	1
same, hyperedge only	−1	1
to_hypergraph() of a four-bar + FixedDyad	8	1
leggedsnake Jansen / Chebyshev / Klann / TrotBot / Strider	17 / 5 / 11 / 34 / 41	1 each
19 catalog entries	1	1 (links unchanged too)

Separately, to_simple_graph() expanded a hyperedge to a chain of N−1 edges while topology.isomorphism and assur.from_hypergraph expand to a clique — so to_mechanism() raised "could not determine solve order" for a ternary link written as a hyperedge alone, unless the caller repeated its edges by hand (which the tutorial told them to do).

The fix

• Links are rigid bodies: the ground (all GROUND nodes), one body per edge and per hyperedge, merged whenever two share ≥ 2 nodes (two links pinned at two points are one link). A node in k bodies is k − 1 one-DOF joints: coupler points are not joints, multiple joints count. A PRISMATIC node is a slider block with a prismatic joint to its guide hyperedge. Three edges closing a triangle without a hyperedge still count as three bars — the DOF is identical, only num_links differs — documented.
• to_simple_graph() expands to the clique, keeping (not duplicating) pairs already joined by an edge. Hyperedge.to_edges() is public and keeps its documented chain.
• Two old tests pinned the wrong values and their own docstrings said so: a lone crank now reports DOF 1 (was −1), a triangle with one side on the ground DOF 0 (was 3). New tests cover the coupler variants, a multiple joint, a slider, the FixedDyad round trip, catalog invariance, and to_mechanism on a hyperedge-only ternary link in every node order.
• Tutorial: counting rules described, the "analyze without the coupler point" caveat replaced by the coupler four-bar's actual mobility.

2608 passed, ruff and mypy clean, the graph tutorial page executes. leggedsnake against this branch: 627 pass, 1 fails — test_triangle_dof_is_three, which pinned the old wrong value (a triangle pinned at one ground node is 1 DOF); to be updated there after release.

Found on the way, not fixed here: RRPDyad → to_hypergraph() → to_mechanism() builds the slider as an RRR dyad — from_sim_linkage only tags LinearActuator nodes PRISMATIC and to_mechanism never sets a slide axis — so a slider-crank round trip raises UnbuildableError a few steps in. Issue to follow.