Add metamorphic dynamics tests and normalize DSL dynamics provenance metadata

[Copilot]

This PR adds a dedicated metamorphic test suite for DSL-driven dynamics simulations to validate invariants under stochastic execution and network transformations, rather than relying on brittle trajectory fixtures. It also aligns dynamics provenance with the metadata contract expected by the issue.

• Metamorphic test coverage (tests/test_dynamics_metamorphic.py)

  • Added compact synthetic fixtures:
    • single-layer 3-node chain
    • two-layer multiplex-like network
    • network with an isolated node
  • Added invariants for DSL Q.dynamics(...) simulations:
    • fixed-seed determinism (SIR, SIS)
    • seed divergence for stochastic runs
    • population conservation via state_counts (SIR, SIS)
    • edge-rate behaviors:
      • beta=0 prevents spread
      • gamma=0 prevents recovery in SIR
      • zero initial infections keeps epidemic at zero
    • multilayer consistency: weighted prevalence_by_layer aggregates to global prevalence (replica semantics explicitly documented)
    • isolated node remains uninfected when not seeded
    • relabeling invariance of aggregate trajectories

• Provenance completeness in DSL dynamics executor

  • Extended execute_dynamics_stmt provenance payload to include:
    • model (name + parameters)
    • randomness (seed)
    • run (steps + replicates)
    • initial_conditions (strategy)
    • network_fingerprint (node/edge/layer counts + layers)
    • py3plex_version
    • backend
  • Preserved existing dynamics config/hash metadata.

Example (new provenance contract asserted in tests):

sim = (
    Q.dynamics("SIR", beta=0.3, gamma=0.1)
    .seed_infections(fraction=0.2)
    .run(steps=15, replicates=4, track=["prevalence"])
    .random_seed(42)
    .execute(net)
)

prov = sim.meta["provenance"]
assert prov["model"]["name"] == "SIR"
assert prov["randomness"]["seed"] == 42
assert prov["run"]["steps"] == 15
assert prov["run"]["replicates"] == 4