Hackathon Project 8: Add CUDA-managed projection container support (continuation of the image stage)

[Dimitra-Kyriakopoulou]

This PR is PART of Hackathon Project 8: update STIR data to CUDA managed pointers, then adapt sirf.STIR data-containers and expose the underlying CUDA managed pointer to Python. It is a continuation of the earlier image stage of the same task. The Stage 1 STIR image part ("images first", via VoxelsOnCartesianGrid for CUDA-managed images) has already been merged upstream in #1693, while the companion Stage 1 SIRF part remains separate in SyneRBI/SIRF#1380.

This new PR contains the STIR continuation for the next two stages of the project. Stage 2 extends the managed-memory path from images to acquisition/projection data, by adding the ProjDataInMemory constructor/storage seam needed for CUDA-managed projection containers. Stage 3 then uses that same STIR continuation as the repository-level basis for the end-to-end managed-memory reconstruction proof. The companion SIRF work is in the fork under Dimitra-Kyriakopoulou/SIRF: the earlier Stage 1 image branch Hackathon_Project8 and the current continuation branch Hackathon_project8_II. The runnable benchmark/proof repository for the whole Project 8 path, including Stage 1, Stage 2, and Stage 3, is Dimitra-Kyriakopoulou/ hackathon_project_08_stir_cuda_managed_pointers.

Changes in this pull request

• Added a ProjDataInMemory constructor that accepts an externally allocated Array<1,float>&& buffer.
• Refactored the projection-data layout setup so the geometry-dependent metadata is initialised correctly for both the old path and the new externally supplied buffer path.
• Scope is deliberately limited to the STIR continuation needed for Stage 2 and then reused by Stage 3.

Testing performed

• Validated that ProjDataInMemory can preserve a managed Array<1,float> buffer.
• Validated the deterministic acquisition benchmark used in Stage 2 through the companion SIRF path.
• Validated the full Stage 3 reconstruction proof through the
  companion SIRF/benchmark path:
  • a real one-subiteration OSMAPOSL update ran with managed acquisition input and managed image output
  • managed and normal reconstructions matched to machine precision
• The runnable proof-of-use benchmarks are kept separately in Dimitra-Kyriakopoulou/hackathon_project_08_stir_cuda_managed_pointers.

Related issues

Checklist before requesting a review

• [] I have performed a self-review of my code
• [] I have added docstrings/doxygen in line with the guidance in the developer guide
• [] I have implemented unit tests that cover any new or modified functionality (if applicable)
• The code builds and runs on my machine
• [] documentation/release_XXX.md has been updated with any functionality change (if applicable)

Contribution Notes

Please tick the following:

• [x ] The content of this Pull Request (the Contribution) is intentionally submitted for inclusion in STIR (the Work) under the terms and conditions of the Apache-2.0 License.
• I (or my institution) have signed the STIR Contribution License Agreement (not required for small changes).

[Dimitra-Kyriakopoulou]

Dear Professor @KrisThielemans ,
THANK YOU WHOLEHEARTEDLY for the comments and the fix!!!
Dimitra