benchmarks: implement repro-collection benchmark with multi-node setup

Now that we can run benchmarks using the multi-node setup in Fastpath, expand repro-collection benchmark to run in 2 separate nodes. In the benchmark's yaml file, provide the roles as sut and loadgen. The nodes configured under selected SUT will be mapped to the roles. Update repro_collection exec.py for running each role process separately on the respective nodes. Frame the command to trigger SUT/LDG workload dynamically and parse the results based on the role configured.

Signed-off-by: Aishwarya Rambhadran aishwarya.rambhadran@arm.com