[CK Profiler] Initialize tensors on GPU in CK profiler

[johannes-graner]

Proposed changes

This PR migrates CK profilers from CPU-side to GPU-side tensor initialization, eliminating a performance bottleneck. The convolution profilers now use the DeviceMemory API methods for on-device data generation, removing expensive CPU-GPU transfers.

Profiler	Before	After	Speedup
Grouped Conv Forward	45.0 s	16.0 s	2.8x
Grouped Conv Backward Data	15.0 s	8.8 s	1.7x
Grouped Conv Backward Weight	20.0 s	6.9 s	2.9x

This saves in total ~50 s per CI run.

Checklist

• I have added tests relevant to the introduced functionality, and the unit tests are passing locally
• I have added the test to REGRESSION_TESTS list defined at the top of CMakeLists.txt in tests/CMakeLists.txt, IF the test takes more than 30 seconds to run.
• I have added inline documentation which enables the maintainers with understanding the motivation
• I have removed the stale documentation which is no longer relevant after this pull request
• (If this change is user-facing) I have added release notes which provide the end users with a brief summary of the improvement from this pull request
• I have run clang-format on all changed files
• Any dependent changes have been merged

[AviralGoelAMD]

Great improvement!