Added rocm lib path discovery with TheRock support

[gahan9]

Motivation

Compilation failure with https://github.com/ROCm/TheRock installation

Technical Details

1. Added path for device libs as to either of generic path $(ROCM_PATH)/amdgcn/bitcode or $(ROCM_PATH)/lib/llvm/amdgcn/bitcode)
2. Replaces Transferbench compilation argument from $(HIPCC) to $(CXX) to comply with overridden CXX compiler environment

Test Plan

1. Install amdgpu driver
2. Install TheRock nightly
3. Install openmpi-bin, openmpi-common and libopenmpi-dev
4. Clone transferbench
5. Modify MPI_PATH to where include/mpi.h exists
6. Modify ROCM_PATH which points to TheRock stack

Test Result

Build successful and execution working as expected:

$ make
Makefile:42: "Could not find /opt/rocm-7.11.0/bin/amdclang++. Using fallback to /opt/rocm-7.11.0/bin/hipcc"
Building with NIC executor support. Can set DISABLE_NIC_EXEC=1 to disable
Building with MPI communicator support.  Can set DISABLE_MPI_COMM=1 to disable
/opt/rocm-7.11.0/bin/hipcc -I/opt/rocm-7.11.0/include -I/opt/rocm-7.11.0/include/hip -I/opt/rocm-7.11.0/include/hsa -DNIC_EXEC_ENABLED -DMPI_COMM_ENABLED -I/usr/lib/x86_64-linux-gnu/openmpi/include -Wall -x hip -D__HIP_PLATFORM_AMD__ -D__HIPCC__ "--offload-arch=native" --rocm-device-lib-path=/opt/rocm-7.11.0/lib/llvm/amdgcn/bitcode -O3 -I./src/header -I./src/client -I./src/client/Presets src/client/Client.cpp -o TransferBench -lnuma -L/opt/rocm-7.11.0/lib -lhsa-runtime64 -lamdhip64 -lpthread -libverbs -L//usr/lib/x86_64-linux-gnu/openmpi/lib -lmpi

$ ./TransferBench 
TransferBench v1.66.00 (with MPI+NIC support) (Single-node mode)
=============================================================================================================
Usage: ./TransferBench config <N>
  config: Either:
          - Filename of configFile containing Transfers to execute (see example.cfg for format)
          - Name of preset config:
  N     : (Optional) Number of bytes to copy per Transfer.
          If not specified, defaults to 268435456 bytes. Must be a multiple of 4 bytes
          If 0 is specified, a range of Ns will be benchmarked
          May append a suffix ('K', 'M', 'G') for kilobytes / megabytes / gigabytes

Environment variables:
======================
 ALWAYS_VALIDATE   - Validate after each iteration instead of once after all iterations
...
...
...

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.