CPP.run(n_jobs>1) crashes with multiprocessing.Manager() EOFError outside notebooks

[breimanntools]

Part of #336 (usability epic).

Problem

CPP.run(..., n_jobs>1) crashes in non-notebook contexts (python -c, heredoc, some subprocess
shells). The progress-sharing path .../cpp/_filters/_progress.py:_get_mp_shared() calls
multiprocessing.Manager(), which raises:

EOFError
  File ".../multiprocessing/managers.py", start -> reader.recv() -> _recv_bytes -> raise EOFError

Workaround: n_jobs=1, or always run from a real .py module file. This makes multi-core CPP fragile
in CI/scripts/pipelines.

Suggestion

Wrap the mp.Manager() spawn in try/except and degrade gracefully to single-process progress
(or a no-op progress bar) instead of aborting the run.