ci: remove dead rm -rf build, add gpu_partition default

- Remove no-op 'rm -rf build' inside 'if [ ! -d build ]' guard
  in common/test.sh and common/bench.sh.
- Default gpu_partition to 'batch' before dynamic selection to
  prevent unbound variable error if a new cluster is added.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: Spencer Bryngelson

.github/scripts/submit-slurm-job.sh

@@ -95,6 +95,7 @@ if [ "$device" = "cpu" ]; then
     esac
 elif [ "$device" = "gpu" ]; then
     # Determine GPU partition
+    gpu_partition="batch"
     if [ "$gpu_partition_dynamic" = "true" ]; then
         # Use pre-selected bench partition if available, otherwise query sinfo
         if [ -n "${BENCH_GPU_PARTITION:-}" ]; then

.github/workflows/common/bench.sh

@@ -23,7 +23,6 @@ fi
 # --- Build (if not pre-built on login node) ---
 # Phoenix builds inside SLURM; Frontier pre-builds via build.sh on the login node.
 if [ ! -d "build" ]; then
-    rm -rf build
     source .github/scripts/retry-build.sh
     retry_build ./mfc.sh build -j $n_jobs $build_opts || exit 1
 fi

.github/workflows/common/test.sh

@@ -11,7 +11,6 @@ build_opts="$gpu_opts"
 # --- Build (if not pre-built on login node) ---
 # Phoenix builds inside SLURM; Frontier pre-builds via build.sh on the login node.
 if [ ! -d "build" ]; then
-    rm -rf build
     source .github/scripts/retry-build.sh
 
     # Phoenix: smoke-test the syscheck binary to catch architecture mismatches