ci: add HF_HOME cache + CUDA_VISIBLE_DEVICES at workflow level

Following the transformers quantization-CI pattern:

- HF_HOME=/mnt/cache: if the runner has persistent storage at this path,
  model downloads are cached across runs. If not, HF hub falls back to
  the default ~/.cache/huggingface.

- CUDA_VISIBLE_DEVICES=0,1 at workflow level: on single-GPU runners,
  device 1 simply doesn't exist so only GPU 0 is used. On multi-GPU
  runners (g6.12xlarge), both are visible. This eliminates per-step env
  overrides.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: Titus-von-Koeller

.github/workflows/tests-integration-nightly.yml

@@ -27,6 +27,8 @@ env:
   TORCH_VERSION: "2.9.1"
   PYPI_INDEX: "https://download.pytorch.org/whl/cu128"
   BNB_WHEEL_URL: "https://github.com/bitsandbytes-foundation/bitsandbytes/releases/download/continuous-release_main/bitsandbytes-1.33.7.preview-py3-none-manylinux_2_24_x86_64.whl"
+  CUDA_VISIBLE_DEVICES: "0,1"
+  HF_HOME: /mnt/cache
 
 jobs:
   # ─── Downstream test jobs ─────────────────────────────────────────────────
@@ -79,7 +81,6 @@ jobs:
         working-directory: /tmp/transformers
         env:
           RUN_SLOW: "1"
-          CUDA_VISIBLE_DEVICES: "0"
         shell: bash -o pipefail {0}
         run: |
           mkdir -p ${GITHUB_WORKSPACE}/reports
@@ -135,7 +136,6 @@ jobs:
         working-directory: /tmp/transformers
         env:
           RUN_SLOW: "1"
-          CUDA_VISIBLE_DEVICES: "0,1"
         shell: bash -o pipefail {0}
         run: |
           mkdir -p ${GITHUB_WORKSPACE}/reports
@@ -191,7 +191,6 @@ jobs:
         working-directory: /tmp/accelerate
         env:
           RUN_SLOW: "1"
-          CUDA_VISIBLE_DEVICES: "0"
         shell: bash -o pipefail {0}
         run: |
           mkdir -p ${GITHUB_WORKSPACE}/reports
@@ -247,7 +246,6 @@ jobs:
         working-directory: /tmp/peft
         env:
           IS_GITHUB_CI: "1"
-          CUDA_VISIBLE_DEVICES: "0"
         shell: bash -o pipefail {0}
         run: |
           mkdir -p ${GITHUB_WORKSPACE}/reports
@@ -305,7 +303,6 @@ jobs:
         working-directory: /tmp/peft
         env:
           IS_GITHUB_CI: "1"
-          CUDA_VISIBLE_DEVICES: "0,1"
         shell: bash -o pipefail {0}
         run: |
           mkdir -p ${GITHUB_WORKSPACE}/reports