confinement: share geometry data between threads

[ManuelHu]

This fixed a very common memory safety error in MT mode with vertex confinement.

ASan report before this fix:

=================================================================
==1149==ERROR: AddressSanitizer: attempting double-free on 0x6020006e8ff0 in thread T2:
    #0 0x55ebc1f0143f in operator delete(void*, unsigned long) (/work/remage/build/src/remage+0x26943f)
    #1 0x7fcd575cf1b2 in G4SolidStore::DeRegister(G4VSolid*) (/opt/geant4/lib/libG4geometry.so+0x1541b2)
    #2 0x7fcd575d1742 in G4VSolid::~G4VSolid() (/opt/geant4/lib/libG4geometry.so+0x156742)
    #3 0x7fcd5763d36d in G4SubtractionSolid::GetCubicVolume() (/opt/geant4/lib/libG4geometry.so+0x1c236d)
    #4 0x7fcd5ef62227 in RMGVertexConfinement::SampleableObject::SampleableObject(G4VPhysicalVolume*, CLHEP::HepRotation, CLHEP::Hep3Vector, G4VSolid*, bool) /work/remage/src/RMGVertexConfinement.cc:61
    #5 0x7fcd5efc205b in void __gnu_cxx::new_allocator<RMGVertexConfinement::SampleableObject>::construct<RMGVertexConfinement::SampleableObject, G4VPhysicalVolume*&, CLHEP::HepRotation const&, CLHEP::Hep3Vector const&, G4VSolid*&, bool&>(RMGVertexConfinement::SampleableObject*, G4VPhysicalVolume*&, CLHEP::HepRotation const&, CLHEP::Hep3Vector const&, G4VSolid*&, bool&) /usr/include/c++/11/ext/new_allocator.h:162
    #6 0x7fcd5efb063d in void std::allocator_traits<std::allocator<RMGVertexConfinement::SampleableObject> >::construct<RMGVertexConfinement::SampleableObject, G4VPhysicalVolume*&, CLHEP::HepRotation const&, CLHEP::Hep3Vector const&, G4VSolid*&, bool&>(std::allocator<RMGVertexConfinement::SampleableObject>&, RMGVertexConfinement::SampleableObject*, G4VPhysicalVolume*&, CLHEP::HepRotation const&, CLHEP::Hep3Vector const&, G4VSolid*&, bool&) /usr/include/c++/11/bits/alloc_traits.h:516
    #7 0x7fcd5efb0bd5 in void std::vector<RMGVertexConfinement::SampleableObject, std::allocator<RMGVertexConfinement::SampleableObject> >::_M_realloc_insert<G4VPhysicalVolume*&, CLHEP::HepRotation const&, CLHEP::Hep3Vector const&, G4VSolid*&, bool&>(__gnu_cxx::__normal_iterator<RMGVertexConfinement::SampleableObject*, std::vector<RMGVertexConfinement::SampleableObject, std::allocator<RMGVertexConfinement::SampleableObject> > >, G4VPhysicalVolume*&, CLHEP::HepRotation const&, CLHEP::Hep3Vector const&, G4VSolid*&, bool&) /usr/include/c++/11/bits/vector.tcc:449
    #8 0x7fcd5ef9d667 in RMGVertexConfinement::SampleableObject& std::vector<RMGVertexConfinement::SampleableObject, std::allocator<RMGVertexConfinement::SampleableObject> >::emplace_back<G4VPhysicalVolume*&, CLHEP::HepRotation const&, CLHEP::Hep3Vector const&, G4VSolid*&, bool&>(G4VPhysicalVolume*&, CLHEP::HepRotation const&, CLHEP::Hep3Vector const&, G4VSolid*&, bool&) /usr/include/c++/11/bits/vector.tcc:121
    #9 0x7fcd5ef64d7b in RMGVertexConfinement::SampleableObjectCollection::emplace_back(G4VPhysicalVolume*, CLHEP::HepRotation const&, CLHEP::Hep3Vector const&, G4VSolid*, bool) /work/remage/src/RMGVertexConfinement.cc:118
    #10 0x7fcd5ef679d4 in RMGVertexConfinement::InitializePhysicalVolumes() /work/remage/src/RMGVertexConfinement.cc:163
    #11 0x7fcd5ef71b29 in RMGVertexConfinement::ActualGenerateVertex(CLHEP::Hep3Vector&) /work/remage/src/RMGVertexConfinement.cc:410
    #12 0x7fcd5ef71474 in RMGVertexConfinement::GenerateVertex(CLHEP::Hep3Vector&) /work/remage/src/RMGVertexConfinement.cc:398
    #13 0x7fcd5eed4d9f in RMGMasterGenerator::GeneratePrimaries(G4Event*) /work/remage/src/RMGMasterGenerator.cc:57
    #14 0x7fcd5a190bfe in G4WorkerTaskRunManager::GenerateEvent(int) (/opt/geant4/lib/libG4run.so+0xc2bfe)
    #15 0x7fcd5a18f093 in G4WorkerTaskRunManager::ProcessOneEvent(int) (/opt/geant4/lib/libG4run.so+0xc1093)
    #16 0x7fcd5a18ed95 in G4WorkerTaskRunManager::DoEventLoop(int, char const*, int) (/opt/geant4/lib/libG4run.so+0xc0d95)
    #17 0x7fcd5a18ef7c in G4WorkerTaskRunManager::DoWork() (/opt/geant4/lib/libG4run.so+0xc0f7c)
    #18 0x7fcd5a1666c1 in std::_Function_handler<std::unique_ptr<std::__future_base::_Result_base, std::__future_base::_Result_base::_Deleter> (), std::__future_base::_Task_setter<std::unique_ptr<std::__future_base::_Result<void>, std::__future_base::_Result_base::_Deleter>, std::__future_base::_Task_state<PTL::TaskGroup<void, void, 0l>::exec<G4TaskRunManager::AddEventTask(int)::{lambda()#1}, , void>(G4TaskRunManager::AddEventTask(int)::{lambda()#1})::{lambda()#1}, std::allocator<int>, void ()>::_M_run()::{lambda()#1}, void> >::_M_invoke(std::_Any_data const&) (/opt/geant4/lib/libG4run.so+0x986c1)
    #19 0x7fcd5a16c3ac in std::__future_base::_State_baseV2::_M_do_set(std::function<std::unique_ptr<std::__future_base::_Result_base, std::__future_base::_Result_base::_Deleter> ()>*, bool*) (/opt/geant4/lib/libG4run.so+0x9e3ac)
    #20 0x7fcd56653ee7 in __pthread_once_slow nptl/pthread_once.c:116
    #21 0x7fcd5a16c521 in PTL::Task<void>::operator()() (/opt/geant4/lib/libG4run.so+0x9e521)
    #22 0x7fcd57161930 in PTL::ThreadPool::execute_thread(PTL::VUserTaskQueue*) (/opt/geant4/lib/libG4ptl.so.2+0x16930)
    #23 0x7fcd571623db in PTL::ThreadPool::start_thread(PTL::ThreadPool*, std::vector<std::shared_ptr<PTL::ThreadData>, std::allocator<std::shared_ptr<PTL::ThreadData> > >*, long) (/opt/geant4/lib/libG4ptl.so.2+0x173db)
    #24 0x7fcd56ffb252  (/lib/x86_64-linux-gnu/libstdc++.so.6+0xdc252)
    #25 0x7fcd5664eac2 in start_thread nptl/pthread_create.c:442
    #26 0x7fcd566dfa03 in __clone (/lib/x86_64-linux-gnu/libc.so.6+0x125a03)

0x6020006e8ff0 is located 0 bytes inside of 8-byte region [0x6020006e8ff0,0x6020006e8ff8)
freed by thread T1 here:
    #0 0x55ebc1f0143f in operator delete(void*, unsigned long) (/work/remage/build/src/remage+0x26943f)
    #1 0x7fcd575cf1b2 in G4SolidStore::DeRegister(G4VSolid*) (/opt/geant4/lib/libG4geometry.so+0x1541b2)

previously allocated by thread T2 here:
    #0 0x55ebc1f003d7 in operator new(unsigned long) (/work/remage/build/src/remage+0x2683d7)
    #1 0x7fcd575cff63 in G4SolidStore::Register(G4VSolid*) (/opt/geant4/lib/libG4geometry.so+0x154f63)

Thread T2 created by T0 here:
    #0 0x55ebc1ea2895 in pthread_create (/work/remage/build/src/remage+0x20a895)
    #1 0x7fcd56ffb328 in std::thread::_M_start_thread(std::unique_ptr<std::thread::_State, std::default_delete<std::thread::_State> >, void (*)()) (/lib/x86_64-linux-gnu/libstdc++.so.6+0xdc328)

Thread T1 created by T0 here:
    #0 0x55ebc1ea2895 in pthread_create (/work/remage/build/src/remage+0x20a895)
    #1 0x7fcd56ffb328 in std::thread::_M_start_thread(std::unique_ptr<std::thread::_State, std::default_delete<std::thread::_State> >, void (*)()) (/lib/x86_64-linux-gnu/libstdc++.so.6+0xdc328)

SUMMARY: AddressSanitizer: double-free (/work/remage/build/src/remage+0x26943f) in operator delete(void*, unsigned long)
==1149==ABORTING

to reproduce

$ cmake -DCMAKE_BUILD_TYPE=ASan ..
$ make
$ export ASAN_OPTIONS=detect_leaks=0:log_path=asan.log
$ ctest -R confinement-mt --repeat-until-fail 100

There are some crashes apart from this, but they also disappeared after this change.

Performance impact: not measured, but waiting time should be negligible compared to total simulation time, as this should only be used once.