run --process/-n accepts 'cpu' to use one subprocess per CPU

[chiliec]

Closes #400.

Implements the -n cpu form that was agreed in the issue thread: --process/-n now accepts cpu in addition to an integer, and resolves it to os.cpu_count(). The default stays 0 (no parallel execution), so nothing changes for existing invocations.

$ doit -n cpu
$ doit -n cpu -P thread

Changes:

• doit/cmd_run.py: num_process_type converter ('cpu' → os.cpu_count() or 1, otherwise int); option type and help text updated.
• doit/cmdparse.py: the "Error parsing parameter" message now prints the type's __name__ instead of its repr, so a callable type renders as (num_process_type) rather than <function ... at 0x...>.
• tests/test_cmd_run.py: test_num_process covering default, integer, cpu, and an invalid value.
• CHANGES, doc/cmd-run.rst.

Validation (Python 3.11, rut 0.4.0):

• rut tests/test_cmd_run.py before the change: test_num_process fails with invalid literal for int() with base 10: 'cpu'; after: 15 passed.
• Full rut: 830 passed, 1 failed — tests/test___main__.py::TestMain.test_execute, which fails identically on a clean checkout of da5a25f here (it imports the repo's dodo.py, whose imports are not installed in my env), so unrelated to this change.
• Manual: doit -n cpu and doit -n 2 -P thread on a two-task dodo.py run both tasks; doit -n xyz exits 3 with Error parsing parameter 'num_process' (num_process_type). expected an integer or 'cpu', got 'xyz'.