fix(copyright): recover contactless POD authors

[mstykow]

Summary

• Recover short contactless names, collectives, and rosters from bounded POD AUTHOR and AUTHORS sections.
• Parse POD =item payloads as individual records and stop at headings, =cut, block boundaries, embedded-document fences, or the existing size cap.
• Split only all-name rosters, detach trailing authored-on dates, and reject role labels and narrative paragraphs.

Issues

• Covers: follow-up author-detection gaps found while validating False-positives when comparing provenant with scancode-toolkit output #1391

Scope and exclusions

• Included: short standalone names, collective credits, conservative comma/and rosters, POD list items, embedded-document fences.
• Explicit exclusions: narrative contribution lists and non-POD source-header credits; those require separate evidence rules.

How to verify

• Scan Perl 5.38.4 with --copyright and compare author values against fix(copyright): recover bounded POD author credits #1410. This branch adds 204 author values across 130 files, removes none, introduces no duplicate values, and leaves copyrights and holders unchanged.
• Scan Info-ZIP 3.0 with --copyright; author, copyright, and holder output should be exactly unchanged from fix(copyright): recover bounded POD author credits #1410.

Intentional differences from Python

• ScanCode has any author output in only 28 affected Perl files and no exact matches for these recovered values. Provenant uses bounded POD structure plus name-shape validation to retain source-faithful names that ScanCode misses.

Follow-up work

• Created or intentionally deferred: recover narrative contribution rosters inside bounded author sections and static source-header credits in separate benchmarked layers.

Expected-output fixture changes

• Files changed: none.
• Why the new expected output is correct: focused end-to-end tests cover the new list, name, collective, date, fence, and rejection behavior without changing existing goldens.

[mstykow]

@greptileai review