ci: shard the test run on mac os X

This should comes with some benefit:
- spread the load across more runner,
- reduce the real-time CI run,
- reduce the "retry" run when we need them.

We start with the Mac jobs, but that would be tremendously useful for Windows
too.

For linux, we need to reduce the startup overhead for this to be worth it.
Building smaller image and speeding up clone should help with that.

contrib/heptapod-ci.yml

@@ -109,6 +109,7 build-c-wheel:
         - PORT_ARG="--port $PORT_START"
         - echo $PORT_ARG
         - WHEEL_ARG=""
+        - SHARDING_ARGS=""
         - if test -n "$WHEEL_TYPE"; then
              PY_TAG=`$PYTHON -c 'import sys; v=sys.version_info; t=f"cp{v.major}{v.minor}"; print(f"{t}-{t}")'`;
              echo "$PY_TAG";
@@ -122,12 +123,18 build-c-wheel:
           else
             echo installing from source;
           fi;
+        - if [ -n "$CI_NODE_INDEX" ]; then 
+            echo "Running the test in multiple shard - [$CI_NODE_INDEX/$CI_NODE_TOTAL]";
+            SHARDING_ARGS="--shard-index $CI_NODE_INDEX --shard-total $CI_NODE_TOTAL";
+            echo "sharding... $SHARDING_ARGS";
+          fi
         - HGTESTS_ALLOW_NETIO="$TEST_HGTESTS_ALLOW_NETIO"
           "$PYTHON" tests/run-tests.py
             --color=always
             $PORT_ARG
             $WHEEL_ARG
             $FLAVOR
+            $SHARDING_ARGS
             $FILTER
             $RUNTEST_ARGS;
 
@@ -309,6 +316,10 windows-pyox:
 macos:
     extends: .test-c
     stage: platform-compat
+    # run the test in multiple shard to help spread the load between concurrent
+    # MR as the macos runner is a shell runner there is not startup overhead
+    # for tests.
+    parallel: 10
     tags:
       - macos
     variables: