Improve C++ coverage for helper utilities and terminal/barrier constraints

[Copilot]

This PR expands coverage in parts of the C++ core that were lightly tested or effectively untested: finite-difference utilities, attitude conversion helpers, and terminal/barrier constraint utilities. It also fixes the terminal-constraint method signatures so those code paths are actually exercised through the existing Constraint interface.

• Finite-difference helper coverage

  • Replaced the minimal finite-difference test with analytic checks for gradient, Jacobian, and Hessian computation.
  • Covers all supported modes: central, forward, and backward differences.
  • Adds explicit coverage for invalid mode handling.

• Attitude conversion helper coverage

  • Added focused tests for Euler/rotation round-trips, gimbal-lock handling, MRP shadow-set equivalence, quaternion-to-MRP edge behavior, and skew-matrix cross-product properties.
  • Exercises previously untested branches in /home/runner/work/cddp-cpp/cddp-cpp/src/cddp_core/helper.cpp.

• Terminal and barrier utility coverage

  • Added tests for terminal equality and inequality constraints, including bounds, Jacobians, Hessian shape/zero structure, violation computation, and dimension mismatch errors.
  • Added tests for relaxed log-barrier value/gradient/Hessian behavior on a linear constraint.
  • Added tests for discrete barrier state initialization, violation extraction, state updates, penalty cost terms, and parameter validation.

• Constraint interface compatibility

  • Updated /home/runner/work/cddp-cpp/cddp-cpp/include/cddp-cpp/cddp_core/terminal_constraint.hpp overrides to match the base Constraint interface signatures, including the optional index parameter.
  • This makes the terminal constraint types compile and behave correctly when used polymorphically.

Example of the new coverage shape:

for (const int mode :
     std::array<int, 3>{kCentralDifference, kForwardDifference,
                        kBackwardDifference}) {
  const Eigen::MatrixXd jac =
      cddp::finite_difference_jacobian(vector_function, x, 1e-6, mode);
  EXPECT_TRUE(jac.isApprox(jac_expected, 1e-4)) << "mode=" << mode;
}