Stabilize new build in separate repository, and get CI/CD working

[CodeByDrescher]

Summary

Stabilize the ODE solver against sanitizers, modernize the build, and ship a
Python distribution alongside the native CLI.

• Build: switch to Conan-managed dependencies (argparse, spdlog,
  libcurl); CMake minimum 3.13 → 3.16; consolidate platform detection on
  WIN32; new options OPTION_TARGET_PYTHON_BINDING,
  OPTION_TEST_WITH_LOCALHOST, OPTION_STATICALLY_LINK; warning cleanup
  pass and sprintf removal so the C++20 build compiles clean.
• Solver architecture: lift input parsing out of the solver into
  VCellSolverInput + VCellSolverFactory; introduce abstract VCellSolver
  base; replace the 273-line hand-rolled argv parser in
  SundialsSolverStandalone with argparse; expose a reusable
  solve(input, output, tid) entry in SundialsSolverInterface (used by
  both the CLI and the Python module).
• Messaging stabilization: decompose SimulationMessaging into
  MessageEventManager (worker thread + queue, with documented lock
  ordering), CurlProxyClasses (AbstractCurlProxy / NullCurlProxy /
  CurlProxy — replaces #ifdef USE_MESSAGING forests with polymorphism),
  JobEventStatus (now a JobEvent::Status namespaced enum), and
  WorkerEvent. Fixes from TSAN/UBSAN/leak runs: data race,
  mutex-access-before-constructor-finished, undefined behavior, memory
  leaks, int overflow in a test. std::jthread → std::thread for
  compiler portability. Net: SimulationMessaging.cpp 755 → 194 lines.
• Python bindings (pyvcell_odesolver): scikit-build-core + vendored
  pybind11; src/main.cpp exposes version() and
  solve(cvode_input_file_path, output_file_path, tid=-1).
• CI/CD: matrix over macOS-15 (arm + intel), Windows, Linux x86_64 +
  arm64, Python 3.10–3.13. Per-platform Conan profiles, cibuildwheel
  Python wheels, lipo macOS Universal build, GHCR manylinux images,
  mold linker on Linux, sanitizer-enabled CMake.
• Tests: replace IDAWin/tests/smoke/smoke.py with a top-level
  tests/ dir of GoogleTest C++ tests (smoke, hello, message-processing)
  plus a test_basic.py that exercises the wheel.

Test plan

• CI green across all 5 platforms × 4 Python versions
• ctest -VV green locally (Linux, macOS); Windows via CI
• SundialsSolverStandalone_x64 runs the smoke .cvodeInput and
  matches *.ida.expected / *.cvode.expected
• pip install of the built wheel exposes
  pyvcell_odesolver.solve(...) and produces output matching the CLI
• Run with -DOPTION_TARGET_MESSAGING=ON against a localhost broker
  (-DOPTION_TEST_WITH_LOCALHOST=ON) — no TSAN/leak warnings
• macOS Universal binary verified with lipo -archs

[CodeByDrescher]

closes #5
closes #7