nvqpp_SampleAsync fails on machines with multiple GPUs

[bmhowe23]

Required prerequisites

• Consult the security policy. If reporting a security vulnerability, do not report the bug using this form. Use the process described in the policy to report the issue.
• Make sure you've read the documentation. Your issue may be addressed there.
• Search the issue tracker to verify that this hasn't already been reported. +1 or comment there if it has.
• If possible, make a PR with a failing test to give us a starting point to work on!

Describe the bug

The nvqpp_SampleAsync test (included in the cuda-quantum repo) fails with the following error: double free or corruption (out)

Steps to reproduce the bug

On a machine with multiple GPUs, build cuda-quantum and then run this command (or something similar):

$ CUDA_VISIBLE_DEVICES=2,3 ctest --test-dir build -R "^nvqpp_SampleAsync$" -V

It will return something like this:

Internal ctest changing into directory: /workspaces/tmp/cuda-quantum/build
UpdateCTestConfiguration  from :/workspaces/tmp/cuda-quantum/build/DartConfiguration.tcl
Parse Config file:/workspaces/tmp/cuda-quantum/build/DartConfiguration.tcl
UpdateCTestConfiguration  from :/workspaces/tmp/cuda-quantum/build/DartConfiguration.tcl
Parse Config file:/workspaces/tmp/cuda-quantum/build/DartConfiguration.tcl
Test project /workspaces/tmp/cuda-quantum/build
Constructing a list of tests
Done constructing a list of tests
Updating test list for fixtures
Added 0 tests to meet fixture requirements
Checking test dependency graph...
Checking test dependency graph end
test 491
    Start 491: nvqpp_SampleAsync

491: Test command: /usr/bin/bash "-c" "rm -f SampleAsync; /workspaces/tmp/cuda-quantum/build/bin/nvq++  --target nvidia-mqpu /workspaces/tmp/cuda-quantum/docs/sphinx/snippets/cpp/using/cudaq/platform/sample_async.cpp -o SampleAsync &&                /workspaces/tmp/cuda-quantum/build/docs/SampleAsync"
491: Working Directory: /workspaces/tmp/cuda-quantum/build/docs
491: Test timeout computed to be: 1500
491: double free or corruption (out)
491: /usr/bin/bash: line 1:  6582 Aborted                 (core dumped) /workspaces/tmp/cuda-quantum/build/docs/SampleAsync
1/1 Test #491: nvqpp_SampleAsync ................***Failed    5.77 sec

0% tests passed, 1 tests failed out of 1

Label Time Summary:
gpu_required    =   5.77 sec*proc (1 test)

Total Test time (real) =   5.78 sec

The following tests FAILED:
        491 - nvqpp_SampleAsync (Failed)
Errors while running CTest
Output from these tests are in: /workspaces/tmp/cuda-quantum/build/Testing/Temporary/LastTest.log
Use "--rerun-failed --output-on-failure" to re-run the failed cases verbosely.

Expected behavior

The test should not fail.

Is this a regression? If it is, put the last known working version (or commit) here.

Yes, this is a regression. git bisect says a8a8ba5 is the problem

Environment

• CUDA Quantum version: a8a8ba5
• Python version:
• C++ compiler:
• Operating system: Ubuntu

Suggestions

No response