feat: let gitlab parallel drive multi-node snakemake

We are using single job slots to run multi-core jobs, which is oversubscribing modes. This MR aims to let gitlab know how many nodes we are using for parallel (or not) snakemake batches without oversubscribing.

benchmarks/ecal_gaps/config.yml

@@ -11,11 +11,12 @@ bench:ecal_gaps:
   stage: benchmarks
   needs:
     - ["sim:ecal_gaps"]
+  parallel: 8
   script:
     - ln -s $LOCAL_DATA_PATH/input input
     - export PYTHONUSERBASE=$LOCAL_DATA_PATH/deps
     - pip install -r benchmarks/ecal_gaps/requirements.txt
-    - snakemake $SNAKEMAKE_FLAGS --cores 8 results/epic_inner_detector/ecal_gaps
+    - snakemake $SNAKEMAKE_FLAGS --cores 1 --batch results/epic_inner_detector/ecal_gaps=$CI_NODE_INDEX/$CI_NODE_TOTAL
 
 collect_results:ecal_gaps:
   extends: .det_benchmark

[Dmitry Kalinkin]

-  parallel: 8
+  parallel: 2

[Dmitry Kalinkin]

-    - snakemake $SNAKEMAKE_FLAGS --cores 1 --batch results/epic_inner_detector/ecal_gaps=$CI_NODE_INDEX/$CI_NODE_TOTAL
+    - snakemake $SNAKEMAKE_FLAGS --cores 5 --batch results/epic_inner_detector/ecal_gaps=$CI_NODE_INDEX/$CI_NODE_TOTAL