Use 4 vCPUs for running coverage-recording tests

We changed the number of vCPUs for most scenarios in 55e5bdf, but
failed to also propagate this to coverage-recording test execution.

.github/workflows/coverage.yaml

@@ -10,7 +10,7 @@ env:
   windows-vcpus: 4
 
 jobs:
-  # This job takes approximately 40 to 75 minutes
+  # This job takes approximately 22 to 75 minutes
   Linux:
     runs-on: ubuntu-20.04
     steps:
@@ -54,7 +54,7 @@ jobs:
           echo "CCACHE_BASEDIR=$PWD" >> $GITHUB_ENV
           echo "CCACHE_DIR=$PWD/.ccache" >> $GITHUB_ENV
       - name: Configure CMake CBMC build with coverage instrumentation parameters
-        run: cmake -S . -Bbuild -Denable_coverage=1 -Dparallel_tests=2 -DCMAKE_CXX_COMPILER=/usr/bin/g++
+        run: cmake -S . -Bbuild -Denable_coverage=1 -Dparallel_tests=${{env.linux-vcpus}} -DCMAKE_CXX_COMPILER=/usr/bin/g++
       - name: Zero ccache stats and limit in size
         run: ccache -z --max-size=7G
       - name: Execute CMake CBMC build