chore(cuda): update cuda-backend with Virtual Memory Manager and NTT refactor (#2127)

Updates `openvm-cuda-backend` with:
- Performance update for better GPU memory usage:
openvm-org/stark-backend#114
- Refactor to avoid some sporadic issues with NTT params initialization:
openvm-org/stark-backend#123

---------

Co-authored-by: Jonathan Wang <[email protected]>

Author: gaxiom

.github/workflows/benchmark-call.yml

@@ -223,6 +223,7 @@ jobs:
           MAX_SEGMENT_LENGTH="--max_segment_length ${{ inputs.max_segment_length }}"
           OUTPUT_PATH="--output_path $(pwd)/$METRIC_PATH"
           echo "INPUT_ARGS=${FEATURES} ${APP_ARG} ${AGG_ARG} ${MAX_SEGMENT_LENGTH} ${OUTPUT_PATH} ${INPUT_ARGS}" >> $GITHUB_ENV
+          echo "VPMM_PAGES=2048" >> $GITHUB_ENV # 4GB
 
       ##########################################################################
       # Find working directory based on benchmark_name and run the benchmark   #

.github/workflows/extension-tests.cuda.yml

@@ -99,6 +99,7 @@ jobs:
           nvcc --version
           echo "NEXTEST_ENV=CUDA_OPT_LEVEL=3" >> $GITHUB_ENV
           echo "FEATURE_ARGS=--features=cuda" >> $GITHUB_ENV
+          echo "VPMM_PAGES=1" >> $GITHUB_ENV
 
       - name: Run extension tests
         env:

.github/workflows/guest-lib-tests.cuda.yml

@@ -34,7 +34,7 @@ jobs:
           - { names: "ruint k256 p256" }
           - { names: "ff_derive pairing" }
     runs-on:
-      - runs-on=${{ github.run_id }}-guest-lib-cuda-${{ github.run_attempt }}-${{ strategy.job-index }}/runner=test-gpu-nvidia/cpu=8+32
+      - runs-on=${{ github.run_id }}-guest-lib-cuda-${{ github.run_attempt }}-${{ strategy.job-index }}/runner=test-gpu-nvidia/cpu=8+32/family=g6+g5+g6e
 
     steps:
       - uses: runs-on/action@v2

.github/workflows/riscv.yml

@@ -55,6 +55,7 @@ jobs:
         run: |
           nvcc --version
           echo "FEATURE_ARGS=--features=cuda,touchemall" >> $GITHUB_ENV
+          echo "VPMM_PAGES=1" >> $GITHUB_ENV
 
       - name: Download and install risc64-elf-gcc
         if: contains(matrix.platform.image, 'x64')