Perftest: Update versions 25.10

Increase perftest vesrion to 6.27
and spec version to 25.10.

Signed-off-by: Shmuel Shaul <sshaul@nvidia.com>

Perftest: Dynamic CUDA linking

This commit refactors the CUDA integration in Perftest by dynamically
loading the CUDA library (`libcuda.so`) instead of linking it
statically.

Changes include:
- Introduced `cuda_loader.c` to handle dynamic loading of CUDA
functions.
- Modified `cuda_memory.c` to use dynamically loaded function pointers
  instead of direct CUDA API calls.
- Ensured proper cleanup of resources by introducing
`unload_cuda_library()`.
- Find CUDA header path automatically and set related defines if exists.

This change increases flexibility, allowing Perftest to be compiled over
systems with cuda and run on both systems with/without CUDA.

Signed-off-by: Shmuel Shaul <sshaul@nvidia.com>

Perftest: changing spec file version to 25.04.0

Signed-off-by: Shmuel Shaul <sshaul@nvidia.com>

Perftest: enable pcie mapping type only if supported by cuda

Enable data direct pcie mapping type only if supported by CUDA
by checking if CU_MEM_RANGE_FLAG_DMA_BUF_MAPPING_TYPE_PCIE exist.

Signed-off-by: Shmuel Shaul <sshaul@nvidia.com>

Perftest: fix qp_timeout with rdma_cm and UD

Upstream commit b8aa202 add a rdma cm qp_timeout support over datapath,
which is not supported over UD qp type.
This commit enable it over RC qp type only.

Signed-off-by: Shmuel Shaul <sshaul@nvidia.com>

perftest: Add minimum rx_depth size in case of SRQ and UD/UC qp type

Add minimum rx_depth size as 100 in SRQ cases with UD/UC qp type so the server
side can have some time to warmup and catch up with the client
side.

Signed-off-by: Shmuel Shaul <sshaul@nvidia.com>

Perftest: changing spec file version to 24.04.0.

Signed-off-by: Hassan Khadour <hkhadour@nvidia.com>

Perftest: Version increase to 6.20

Signed-off-by: Hassan Khadour <hkhadour@nvidia.com>

Merge pull request #224 from changchengx/flow_label

perftest: support set flow_label through env variable FLOW_LABEL

Perftest: Fix and optimize initial communication functions.

Fix and optimize some code sections in initial communication
functions.

Signed-off-by: Hassan Khadour <hkhadour@nvidia.com>