switch to Ninja+ccache, add clang/lld fast build profile

- CMakePresets.json: add 'fast' preset (clang+lld+ccache)
- .gitignore: generalize build/ to build*/, add CMakeUserPresets.json
- CMakeUserPresets.json: untrack (conan-generated, now gitignored)
- src/Core/Archives.h: remove redundant 'using basic_xml_iarchive::load_override'
  in xml_iarchive; caused ambiguous overload with clang (diamond inheritance)
- src/Core/Object.cpp: remove invalid explicit instantiations of non-template
  virtual Object::serialize (GCC extension, clang rejects)
- README.md, CLAUDE.md: document GCC and LLVM/clang build workflows

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

CLAUDE.md

@@ -12,7 +12,7 @@ export MAMBA_ROOT_PREFIX="/home/share/micromamba"
 eval "$(/home/share/micromamba/bin/micromamba shell hook --shell bash)"
 micromamba activate mutom
 
-# Configure (from repo root, using Conan preset)
+# Configure (from repo root, using Conan preset — uses Ninja + ccache)
 cmake --preset conan-release
 
 # Build everything
@@ -40,6 +40,18 @@ conan install . --output-folder=build --build=missing
 cmake --preset conan-release
 ```
 
+### Build acceleration (already configured)
+- **Ninja** generator — used automatically via the conan default profile (`~/.conan2/profiles/default`)
+- **ccache** — enabled via `CMAKE_CXX_COMPILER_LAUNCHER=ccache`; cached rebuilds are nearly instant (~0.3s vs ~25s cold)
+- **Clang 22 + lld** profile available (`~/.conan2/profiles/fast`) but blocked by template overload ambiguities in `src/Core/Archives.h` that need fixing for full compatibility
+
+To reconfigure with the fast profile once Archives.h is fixed:
+```bash
+conan install . --output-folder=build --build=missing --profile=fast
+cmake -B build -G Ninja -DCMAKE_TOOLCHAIN_FILE=build/conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Release
+cmake --build build -j$(nproc)
+```
+
 ## Architecture
 
 **uLib** is a C++ framework for Cosmic Muon Tomography (CMT), structured as layered shared libraries: