Support Turbomole's multi-file output in read_cclib/cclib_thermo

[galjos]

Found by fetching a real Turbomole calculation from cclib's own public regression-test data (cclib/cclib-data) and running it through cclib_thermo — the first genuine (non-mocked) Turbomole test this reader has ever had.

The bug

Turbomole splits a job's output across many small files instead of one logfile (control, coord, aoforce.out, ...). cclib's own ccread/ccopen documents exactly this case — source: str | list[str], "a single logfile, a list of logfiles (for a single job)" — but read_cclib/cclib_thermo always did cclib.io.ccread(str(path)), which converts a list to its string repr rather than passing it through. So despite Turbomole being named in the docs and the module docstring as a supported program, it could not actually be read at all.

The fix

read_cclib(path) / cclib_thermo(output_file, ...) now accept either a single path or a list of paths (_normalize_source), passed to cclib unmodified for the list case, with a readable _describe_source for error messages either way. Single-file callers (Gaussian/ORCA/Psi4/NWChem) are unaffected.

Verification

Downloaded a real Turbomole 7.2 aoforce frequency calculation (a 7-atom Au/N/Cl complex) from cclib's own public test-data repository and ran it through the actual code:

INFO:cclib:Identified logfile to be in Turbomole format
symbols: ['Cl', 'Au', 'N', 'N', 'Au', 'N', 'N']
15 frequencies (3*7-6), all real — a genuine minimum
electronic_energy(): -951.0820621320888 Hartree
entropy: 110.13 cal/(mol*K)  — physically sensible for this heavier, larger complex

Committed the real captured files as test fixtures (tests/data/calculator/turbomole/, ~160 KB, with a provenance README) and added tests: exact-value parsing, the full cclib_thermo pipeline end-to-end, a mock proving the list reaches cclib.io.ccread unmodified, and unit tests for the new helpers. Full suite passes (377 passed, 12 pre-existing skips), 100% patch coverage on qm.py. Docs updated with a Turbomole multi-file example.

Closes #109

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.28%. Comparing base (bcb49f1) to head (467d002).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #110      +/-   ##
==========================================
+ Coverage   97.27%   97.28%   +0.01%     
==========================================
  Files          31       31              
  Lines        1910     1918       +8     
==========================================
+ Hits         1858     1866       +8     
  Misses         52       52              

Flag	Coverage Δ
unittests	97.28% <100.00%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.