Updated Python executable path in NIXL KVBench workloads

[Bohatchuk]

Summary

Due to the changes implemented in NIXLBench container (ai-dynamo/nixl#845) (Python venv is part of PATH) and to support further integration compatibility with NIXLBench container updated Python executable path for NIXL KVBench and NIXL KVBench CTPerf workloads.

Test Plan

Real-platform runs of NIXL KVBench CTPerf and NIXL KVBench w/ POSIX and UCX backends on internal clusters (both aarch64 and x86_64 archs):

• uv run cloudai run --system-config conf/system/system.toml --test-scenario conf/test_scenario/kvbench_ctperf.toml

name = "NIXL.KVBench.CTPerf"

[[Tests]]
id = "NIXL.KVBench.CTPerf"
name = "KVBenchCTPerf"
description = "NIXL KVBench CTPerf"
test_template_name = "NixlPerftest"
num_nodes = 2
time_limit = "00:45:00"

  [Tests.cmd_args]
  docker_image_url = "urm.nvidia.com#.../nixlbench:...-latest"
  subtest = "sequential-ct-perftest"
  num_user_requests = 10
  batch_size = 1
  num_prefill_nodes = 1
  num_decode_nodes = 1
  prefill_tp = 4
  decode_tp = 4
  isl_mean = 10000
  isl_scale = 3000
  model = "deepseek-r1-distill-llama-70b"

• uv run cloudai run --system-config conf/system/system.toml --test-scenario conf/test_scenario/kvbench_posix.toml

name = "NIXL.KVBench.POSIX"

[[Tests]]
id = "NIXL.KVBench.POSIX"
name = "KVBenchProfile"
description = "NIXL KVBench w/ POSIX backend"
test_template_name = "NIXLKVBench"
num_nodes = 1
time_limit = "00:10:00"

  [Tests.cmd_args]
  docker_image_url = "urm.nvidia.com#.../nixlbench:...-latest"
  model = "./examples/model_deepseek_r1.yaml"
  model_config = "./examples/block-tp1-pp16.yaml"
  backend = "POSIX"
  num_requests = 1
  source = "file"
  num_iter = 16
  page_size = 256
  filepath = "/data"

• uv run cloudai run --system-config conf/system/system.toml --test-scenario conf/test_scenario/kvbench_ucx.toml

name = "NIXL.KVBench.UCX"

[[Tests]]
id = "NIXL.KVBench.UCX"
name = "KVBenchProfile"
description = "NIXL KVBench w/ UCX backend"
test_template_name = "NIXLKVBench"
num_nodes = 2
time_limit = "00:10:00"

  [Tests.cmd_args]
  docker_image_url = "urm.nvidia.com#.../nixlbench:...-latest"
  model = "./examples/model_deepseek_r1.yaml"
  model_config = ["./examples/layer-tp1-pp1.yaml", "./examples/block-tp1-pp16.yaml"] 
  backend = "UCX"
  source = ["memory", "gpu"]
  op_type = ["READ", "WRITE"]