Bug: 35/46 tests fail on ROCm when using address sanitizers.

Description how to reproduce the bug

cmake -DRESOLVE_USE_ASAN=yes -DRESOLVE_USE_UBSAN=yes -DCMAKE_BUILD_TYPE=RelWithDebInfo -DRESOLVE_USE_LUSOL=yes  -DRESOLVE_USE_KLU=yes -DRESOLVE_USE_HIP=yes -DCMAKE_INSTALL_PREFIX=../install ../ReSolve/
make -j
ctest -j

Memory leaks such as:

Indirect leak of 192 byte(s) in 2 object(s) allocated from:
    #0 0x00000035618d  (/autofs/nccs-svm1_home2/regevs/ReSolve_dir/build/tests/functionality/sys_rand_gmres_test.exe+0x35618d)
    #1 0x7fcac46ded82  (/opt/rocm-6.4.0/lib/llvm/bin/../../../lib/librocsparse.so.1+0xc21d82) (BuildId: 5343bf0c81290fba73afda9c5bcff0800301e5e2)

Indirect leak of 64 byte(s) in 2 object(s) allocated from:
#0 0x00000035618d (/autofs/nccs-svm1_home2/r

The same errors do not appear when the address sanitizer is not enabled, or alternatively during CPU only builds.

### Re::Solve version

All versions

### System and environment details

Build environment:
  - Frontier
  - rocm 6.4
  - gcc 12.3

### Additional information

_Verified as AMD's error._

Reproduce the issue with the provided CMake options on Frontier using ROCm 6.4, then run ctest -j and inspect the failures, including functionality/sys_rand_gmres_test.exe. Determine whether the reported sanitizer leaks originate in ReSolve or the ROCm libraries; done requires resolving or clearly isolating the 35 failing tests.