Replace diff_engine_core submodule with subtree

[PTNobel]

Description

Replace the diff_engine_core git submodule with a git subtree. This eliminates the need for git submodule init/update when cloning the repo — the code is included directly.

The subtree is squashed from https://github.com/dance858/DNLP-diff-engine at commit fa65481.

Future updates can be pulled with:

git subtree pull --prefix=diff_engine_core https://github.com/dance858/DNLP-diff-engine <branch> --squash

Type of change

• New feature (backwards compatible)
• New feature (breaking API changes)
• Bug fix
• Other (Documentation, CI, ...)

Contribution checklist

• Add our license to new files.
• Check that your code adheres to our coding style.
• Write unittests.
• Run the unittests and check that they're passing.
• Run the benchmarks to make sure your change doesn't introduce a regression.

🤖 Generated with Claude Code

[PTNobel]

I think Daniel and William will have to update the copyright headers in the DNLP engine before this gets merged into CVXPY master.

[github-actions]

Benchmarks that have stayed the same:

   before           after         ratio
 [a87c4f1f]       [9a084cf4]
      487±0ms          512±0ms     1.05  semidefinite_programming.SemidefiniteProgramming.time_compile_problem
      2.80±0s          2.94±0s     1.05  quantum_hilbert_matrix.QuantumHilbertMatrix.time_compile_problem
     13.7±0ms         13.9±0ms     1.02  simple_LP_benchmarks.SimpleFullyParametrizedLPBenchmark.time_compile_problem
      1.57±0s          1.59±0s     1.01  tv_inpainting.TvInpainting.time_compile_problem
     14.3±0ms         14.4±0ms     1.00  simple_QP_benchmarks.ParametrizedQPBenchmark.time_compile_problem
      1.00±0s          1.01±0s     1.00  finance.FactorCovarianceModel.time_compile_problem
      140±0ms          141±0ms     1.00  high_dim_convex_plasticity.ConvexPlasticity.time_compile_problem
      1.82±0s          1.83±0s     1.00  simple_QP_benchmarks.UnconstrainedQP.time_compile_problem
      12.7±0s          12.7±0s     1.00  finance.CVaRBenchmark.time_compile_problem
      4.17±0s          4.16±0s     1.00  huber_regression.HuberRegression.time_compile_problem
      272±0ms          272±0ms     1.00  matrix_stuffing.ParamSmallMatrixStuffing.time_compile_problem
      4.93±0s          4.92±0s     1.00  optimal_advertising.OptimalAdvertising.time_compile_problem
     38.1±0ms         38.0±0ms     1.00  matrix_stuffing.SmallMatrixStuffing.time_compile_problem
      306±0ms          305±0ms     1.00  gini_portfolio.Yitzhaki.time_compile_problem
      1.37±0s          1.36±0s     1.00  matrix_stuffing.ParamConeMatrixStuffing.time_compile_problem
      21.8±0s          21.7±0s     1.00  sdp_segfault_1132_benchmark.SDPSegfault1132Benchmark.time_compile_problem
      233±0ms          233±0ms     1.00  simple_QP_benchmarks.SimpleQPBenchmark.time_compile_problem
      876±0ms          873±0ms     1.00  simple_LP_benchmarks.SimpleScalarParametrizedLPBenchmark.time_compile_problem
      10.9±0s          10.9±0s     1.00  simple_LP_benchmarks.SimpleLPBenchmark.time_compile_problem
      4.80±0s          4.77±0s     0.99  svm_l1_regularization.SVMWithL1Regularization.time_compile_problem
      676±0ms          672±0ms     0.99  matrix_stuffing.ConeMatrixStuffingBench.time_compile_problem
      228±0ms          226±0ms     0.99  gini_portfolio.Murray.time_compile_problem
      271±0ms          269±0ms     0.99  slow_pruning_1668_benchmark.SlowPruningBenchmark.time_compile_problem
      941±0ms          932±0ms     0.99  gini_portfolio.Cajas.time_compile_problem
      800±0ms          791±0ms     0.99  simple_QP_benchmarks.LeastSquares.time_compile_problem

[Transurgeon]

thanks for making this PR @PTNobel , as discussed offline we ended up creating a PyPI package called sparsediffpy which exposes the bindings there :). Will update the cvxpy PR soon!