[sap] Avoid factoring ununitialized memory

[jwnimmer-tri]

Hotfix for #21421:

[ RUN      ] DenseSuperNodalSolver.InterfaceTest
==20== Conditional jump or move depends on uninitialised value(s)
==20==    at 0x14A588: Eigen::LDLT<Eigen::Ref<Eigen::Matrix<double, -1, -1, 0, -1, -1>, 0, Eigen::OuterStride<-1> >, 1>& Eigen::LDLT<Eigen::Ref<Eigen::Matrix<double, -1, -1, 0, -1, -1>, 0, Eigen::OuterStride<-1> >, 1>::compute<Eigen::Matrix<double, -1, -1, 0, -1, -1> >(Eigen::EigenBase<Eigen::Matrix<double, -1, -1, 0, -1, -1> > const&) (LDLT.h:517)
==20==    by 0x14A1A7: Eigen::LDLT<Eigen::Ref<Eigen::Matrix<double, -1, -1, 0, -1, -1>, 0, Eigen::OuterStride<-1> >, 1>::LDLT<Eigen::Matrix<double, -1, -1, 0, -1, -1> >(Eigen::EigenBase<Eigen::Matrix<double, -1, -1, 0, -1, -1> >&) (LDLT.h:137)
==20==    by 0x149A53: drake::multibody::contact_solvers::internal::DenseSuperNodalSolver::DenseSuperNodalSolver(std::vector<Eigen::Matrix<double, -1, -1, 0, -1, -1>, std::allocator<Eigen::Matrix<double, -1, -1, 0, -1, -1> > > const*, drake::multibody::contact_solvers::internal::BlockSparseMatrix<double> const*) (dense_supernodal_solver.cc:29)
==20==    by 0x11D5B2: drake::multibody::contact_solvers::internal::DenseSuperNodalSolver_InterfaceTest_Test::TestBody() (dense_supernodal_solver_test.cc:129)
...

---

This change is 

[jwnimmer-tri]

@drake-jenkins-bot linux-jammy-clang-bazel-experimental-valgrind-memcheck please

[jwnimmer-tri]

+@sherm1 for both reviews, please (follow-up from #21421).

\CC @amcastro-tri FYI

[jwnimmer-tri]

Reviewable status: LGTM missing from assignee sherm1(platform)

---

multibody/contact_solvers/sap/dense_supernodal_solver.cc line 10 at r1 (raw file):

namespace internal {

int SafeGetCols(const BlockSparseMatrix<double>* J) {

FYI This dance was not necessary. The line J_(SafeDeference("J", J)), already proved that J was non-null; we don't need to re-check before calling J->cols().

[jwnimmer-tri]

@drake-jenkins-bot linux-jammy-clang-bazel-experimental-release please (flake)

[sherm1]

x 2
Thanks, Jeremy!

Reviewed 1 of 1 files at r1, all commit messages.
Reviewable status: complete! all discussions resolved, LGTM from assignee sherm1(platform)

---

multibody/contact_solvers/sap/dense_supernodal_solver.cc line 23 at r1 (raw file):

    : A_(SafeDeference("A", A)),
      J_(SafeDeference("J", J)),
      H_(Eigen::MatrixXd::Zero(J->cols(), J->cols())),

FYI attn @amcastro-tri
Is there any reason you want to leave H_ uninitialized? If so you'll need to rework this so it doesn't look at any of that uninitialized memory. For now we're going to set it to zero to fix CI but you can revisit later if it's important.

[sherm1]

CI failure looks like an unrelated flake?

Reviewable status: complete! all discussions resolved, LGTM from assignee sherm1(platform)

[jwnimmer-tri]

@drake-jenkins-bot linux-jammy-clang-bazel-experimental-everything-release please (flake)