Fix CUDA architecture selection and optimize LBVH Morton codes

[zfergus]

Description

Two independent small fixes, split out of the CUDA branch so they can land on their own.

• tests/CMakeLists.txt — the test target hard-coded CUDA_ARCHITECTURES "native", which makes CMake probe for a physical GPU. Configuring therefore fails outright in GPU-less environments (CI runners, Docker image builds), even when the caller already passed -DCMAKE_CUDA_ARCHITECTURES. We now use whatever the user supplied and fall back to "native" only when they supplied nothing.
• src/ipc/broad_phase/lbvh.cpp — LBVH::init_bvh divided by mesh_width inside the tbb::parallel_for over boxes. Hoisting 1 / mesh_width out of the loop turns three divides per box into three multiplies.

Behavior

• The CMake change is inert for anyone who already passes CMAKE_CUDA_ARCHITECTURES, and unchanged for anyone building on the machine they target.
• The Morton hoist can differ in the last bit, since a * (1/b) does not always round to a / b. At worst that shifts the Morton sort order for a near-tie, changing the tree shape. The candidate set the query returns is unaffected — the traversal is exhaustive over overlapping nodes, not order-dependent.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• Enhancement (non-breaking change which improves existing functionality)

How Has This Been Tested?

• [broad_phase] ~[!benchmark] — 13 test cases, 1,464,038 assertions, all passing. Covers LBVH::detect_*_candidates and Compare BP against brute force, which checks the LBVH candidate set against brute force.
• Full ipc_toolkit, ipc_toolkit_tests, and ipctk builds.

The CMake change is not covered: this was a IPC_TOOLKIT_WITH_CUDA=OFF build, so the edited branch never executed. It needs a CUDA configure to exercise.

Test Configuration:

• OS and Version: macOS 26.6 (arm64)
• Compiler and Version: AppleClang 21.0, CMAKE_BUILD_TYPE=Release, IPC_TOOLKIT_BUILD_TESTS=ON, IPC_TOOLKIT_WITH_CUDA=OFF

Checklist

• I have followed the project style guide
• My code follows the clang-format style guidelines of this project
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• Any dependent changes have been merged and published in downstream modules

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.70%. Comparing base (b778f64) to head (edfe799).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #252   +/-   ##
=======================================
  Coverage   96.70%   96.70%           
=======================================
  Files         191      191           
  Lines       17292    17293    +1     
  Branches      928      927    -1     
=======================================
+ Hits        16722    16723    +1     
  Misses        570      570           

Flag	Coverage Δ
unittests	96.70% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.