feat(graph): add production graph metric lifecycle and queries

.github/workflows/zig-tests.yml

@@ -407,9 +407,10 @@ jobs:
         run: python3 scripts/packaging/test_reproducible_tar.py
 
       # ARC may expose the shared node's memory total when the runner cgroup is
-      # unlimited. Cap the aggregate scheduler budget at 20 GiB so concurrent
+      # unlimited. Cap the aggregate scheduler budget at 22 GiB so concurrent
       # test runtimes cannot overcommit the runner while still admitting the
-      # largest individual compile step, which reserves 20 GiB. The patched
+      # largest individual compile step, which reserves 22 GiB. Cgroup detection
+      # still reserves 20% headroom and never raises a smaller host's budget. The patched
       # 0.16 runner correctly reserves each ready step's max_rss claim as it
       # wakes it.
       - name: Configure bounded parallel unit tests
@@ -419,7 +420,7 @@ jobs:
           jobs=$(nproc)
           if [ "$jobs" -gt 8 ]; then jobs=8; fi
           echo "list=0-$((jobs - 1))" >> "$GITHUB_OUTPUT"
-          echo "max_rss=$(python3 zig/tools/run_bounded_zig_build.py --print-max-rss --max-rss-cap 21474836480)" >> "$GITHUB_OUTPUT"
+          echo "max_rss=$(python3 zig/tools/run_bounded_zig_build.py --print-max-rss --max-rss-cap 23622320128)" >> "$GITHUB_OUTPUT"
           zig_lib_dir="$(zig env | sed -n 's/^[[:space:]]*\.lib_dir = "\(.*\)",$/\1/p')"
           patched_runner="${RUNNER_TEMP}/zig-build-runner-maxrss.zig"
           python3 zig/tools/patch_zig_0_16_build_runner_maxrss.py \
@@ -669,14 +670,14 @@ jobs:
           python3 -m unittest discover -s scripts/bench/pdf -p 'test_*.py'
 
       # Keep normal build-step parallelism while bounding aggregate compiler
-      # memory to the pod's cgroup budget.
+      # memory to the pod's cgroup budget and the 22 GiB largest-step claim.
       - name: Configure bounded parallel heavy tests
         id: cpus
         run: |
           jobs=$(nproc)
           if [ "$jobs" -gt 8 ]; then jobs=8; fi
           echo "list=0-$((jobs - 1))" >> "$GITHUB_OUTPUT"
-          echo "max_rss=$(python3 zig/tools/run_bounded_zig_build.py --print-max-rss --max-rss-cap 21474836480)" >> "$GITHUB_OUTPUT"
+          echo "max_rss=$(python3 zig/tools/run_bounded_zig_build.py --print-max-rss --max-rss-cap 23622320128)" >> "$GITHUB_OUTPUT"
           zig_lib_dir="$(zig env | sed -n 's/^[[:space:]]*\.lib_dir = "\(.*\)",$/\1/p')"
           patched_runner="${RUNNER_TEMP}/zig-build-runner-maxrss.zig"
           python3 zig/tools/patch_zig_0_16_build_runner_maxrss.py \