[Sofa.Simulation.Core] MappingGraph: fix memory leak in from parent/child shared_ptr cycle

[fredroy]

From Claude (sonnet 😅):

BaseMappingGraphNode stored both m_parents and m_children as owning shared_ptrs. MappingGraph::addEdge() linked every connected pair in both directions, creating a reference cycle between each parent and child node.
MappingGraph::clear() (called at the start of every build()) only drops the graph's own m_allNodes vector; it never breaks these cycles. Since EulerExplicitSolver::solve() and EulerImplicitSolver::solve() call
m_mappingGraph.build() unconditionally on every timestep, the entire previous graph leaked on every single animation step for as long as the solver was active.
Fix: store m_parents as std::weak_ptr instead of shared_ptr. Node ownership already belongs to MappingGraph::m_allNodes (and to each node's owning parent via m_children), so the back-reference doesn't need to be owning.

Solver step	Live nodes (before fix)	Live nodes (after fix)
1	0	0
101	1200	24
1001	12000	24
1901	22800	24

---

Basically the shared_ptr retained a ref so the parents/childs were owning each other, so the shared_ptr/gc mechanism were effectively dismissed.

Fix #6230

[with-all-tests]

---

By submitting this pull request, I acknowledge that
I have read, understand, and agree SOFA Developer Certificate of Origin (DCO).

---

Reviewers will merge this pull-request only if

• it builds with SUCCESS for all platforms on the CI.
• it does not generate new warnings.
• it does not generate new unit test failures.
• it does not generate new scene test failures.
• it does not break API compatibility.
• it is more than 1 week old (or has fast-merge label).

[fredroy]

for windows
[ci-build][with-all-tests]

[alxbilger]

Should the children be weak pointers too ? They are supposed to be in m_allNodes too.