fix(copyright): recover bounded POD contributor rosters

[mstykow]

Summary

• Recover bounded contributor rosters from narrative sentences inside POD AUTHOR and AUTHORS sections.
• Support contribution, thanks, supplied-by, attribution, and subject-provided cues without treating unrelated prose as names.
• Preserve valid name particles and contributor collectives while trimming dates, roles, and non-person continuations.

Issues

• Covers: follow-up findings from False-positives when comparing provenant with scancode-toolkit output #1391

Scope and exclusions

• Included: cue-backed narrative credits inside structurally bounded POD author sections.
• Explicit exclusions: unbounded prose, operational contacts, copyright-only contacts, and generic contribution objects.

How to verify

• Scan Perl 5.38.4 with provenant --copyright --json out.json perl-5.38.4. Relative to fix(copyright): recover contactless POD authors #1411, this adds 57 valid author values across 15 files, removes none, introduces no duplicates, and leaves copyright and holder output unchanged.
• Compare the same 15 files with ScanCode: ScanCode reports any author in nine and exactly matches none of the recovered values.
• Scan InfoZIP 3.0 as a control: author, copyright, and holder output is byte-for-byte unchanged from fix(copyright): recover contactless POD authors #1411.

Intentional differences from Python

• Provenant treats bounded, grammatically cued contributor sentences as authorship evidence. ScanCode generally misses these rosters, while Provenant keeps extraction section-bounded and rejects generic prose objects.

Follow-up work

• Created or intentionally deferred: source-header credits and malformed legacy author cleanup remain in later stacked branches.

Expected-output fixture changes

• Files changed: none.
• Why the new expected output is correct: focused detector tests cover cue families, name particles, collectives, exclusions, and section boundaries; the Perl and InfoZIP comparisons provide positive and negative corpus evidence.

[mstykow]

@greptileai review