Add MVP ELN integration for eLabFTW and Kadi4Mat

[linuxrider]

Introduce unitpackage.eln with a shared BaseELNClient ABC and two backend implementations (ElabFTWClient, KadiClient) that support fetch_entry, fetch_entries, upload_entry, and info operations.

Includes Click CLI (unitpackage elabftw/kadi), TOML-based config profiles, fixture recording infrastructure for replay-based tests, and recorded fixtures from eLabFTW 5.5.14 and Kadi4Mat 1.10.0.

Checklist

• Added an entry in doc/news/.
• Added a test for this change.
• Adapted the documentation for this change.

[CAOShurong]

Independent interoperability check at head 424c5b21005a9fbfa4a7c4ea7eb5c21a4c6c1932 (original base 1332b8907f7e5eb106ea9e0c29958bad56b4a72a):

• On Windows/Python 3.12 with elabapi-python 5.6.1 and kadi-apy 0.51.0, I replayed the checked-in, GPL-3.0-or-later fixtures from eLabFTW 5.5.14 and Kadi4Mat 1.10.0 through the two actual fetch_entry implementations, without network calls or external archives.
• The resulting Entry objects were equivalent across the DataFrame, complete field descriptors, resource metadata, and identifier: 3 rows; columns t,E,j; units s,V,A / m2; identifier fixture_test; citation key fixture_test_2026.
• Both reconstructed CSVs and the recorded source payload have SHA-256 e2660b35fbdf08ca162947aa9fd72fd1d1c6f8e1bd4c6cdb1c52f3e4d87b35e7.
• Focused recorded-response tests: 25 passed. Full pytest -q --doctest-modules unitpackage: 340 passed, 3 failed; the three failures are the same Windows config_path assertions already visible in Actions, so I did not treat them as a new interoperability finding. The pinned base ran 149 passed and does not contain unitpackage.eln.

Minimal cross-backend comparison using the PR's fixture wiring:

import pandas as pd
from unitpackage.test.eln.test_elabftw import TestFixtureVersions as ElabReplay
from unitpackage.test.eln.test_kadi import TestFixtureVersions as KadiReplay

def field_map(entry):
    values = [f.to_dict() if hasattr(f, "to_dict") else f for f in entry.fields]
    return {f["name"]: f for f in values}

ec, eid = ElabReplay()._wire_client("5.5.14")
kc, record = KadiReplay()._wire_client("1.10.0")
e = ec.fetch_entry(eid)
k = kc.fetch_entry(record.id)
pd.testing.assert_frame_equal(e.df, k.df)
assert field_map(e) == field_map(k)
assert e.resource.custom["metadata"] == k.resource.custom["metadata"]
assert e.identifier == k.identifier == "fixture_test"