zephyr-cp: run the tests in parallel

The suite runs one test at a time and is 17 of the zephyr-tests job's
34 to 44 minutes. 84% of that is bsim, where every test starts a phy
and one or two simulated devices, and the asan board from adafruit#11317
doubled the native tests, 220 -> 322.

Run pytest with one worker per CPU. --dist loadfile keeps each file on
one worker, which test_web_workflow.py needs for its fixed port 8090;
the rest is already per test, tmp_path for the flash images and a sim
id from the test's node id.

Measured on four cores, whole suite, 322 passed each time: 758 s with
one worker, 350 s with two, 249 s with three, 224 s with four. -n auto
counts physical cores, so it picks 2 on the runner and the host's count
inside a container. sysctl is the macOS spelling of nproc, as in
tools/ci.sh.

Author: lynt-smitka

ports/zephyr-cp/Makefile

@@ -141,4 +141,4 @@ $(foreach board,$(TEST_BOARDS),$(eval $(call test_board_rule,$(board))))
 
 test: $(TEST_BOARDS:%=test-build-%)
 	pytest cptools/tests
-	pytest tests/ -v
+	pytest tests/ -v -n $$(nproc 2>/dev/null || sysctl -n hw.ncpu) --dist loadfile

requirements-dev.txt

@@ -46,4 +46,5 @@ setuptools
 tomlkit
 pytest
 pytest-rerunfailures
+pytest-xdist
 perfetto