[shimwell]

Purpose

Prerequisite for #41 and #42 and the upcoming JEFF generation PR. Merge this first; rebase the feature PRs afterward. No nuclear data processing code or library sources are changed.

Changes

• Run deterministic offline tests, installed CLI help checks, and validation of all URL registry entries on pull requests and pushes.
• Keep strict external URL probes in a separate nightly/manual job. Full multi-GB processing is manual with an explicit script choice, rather than automatically triggered by shared utility changes.
• Explicitly skip only https://nds.iaea.org/fendl/data/neutron/fendl-FENDL-3.2c-neutron-endf.zip and report it as not tested. Other HTTP errors, including Cloudflare challenges, remain failures.
• Keep tests in the existing flat tests directory, register network/integration markers, and ignore local pytest scratch files.

Validation

• 228 passed, 1 environment-dependent skip with OpenMC; 229 passed without OpenMC.
• Installed-package URL tests: 226 passed. All 24 existing CLI help checks passed.
• Targeted live checks: one HDF5 success and the one explicit IAEA skip.
• Workflow structure and preserved manual processing recipes checked; editor diagnostics and diff checks passed.

Coverage boundary

Green PR checks do not establish that remote archives are available. Nightly/manual health checks can remain red for blocked IAEA or obsolete OECD links; these failures are not converted into successes. Full processing was not rerun.

All commits and this PR are attributed to shimwell.

[shimwell]

Verified CI on 9ffc9e0: test_existence/run_scripts passed (5m28s), test_urls/ping_urls passed (23s). live_urls was skipped as designed for a pull-request event; nightly/manual HTTP availability is not claimed to be passing. PR and commit author/committer are shimwell. Ready for review and merge before rebasing #41, #42, and the prepared generation work.