fix(copyright): recover contact-backed action credits

[mstykow]

Summary

• recover explicit contact-backed action credits such as Configured by, Improved by, Originally contributed by, language-specific -ified by, and artifact-role by lines
• normalize bracketed contacts and bound trailing dates, project clauses, and chained attributions
• reject configured machine-account placeholders after consuming aligned punctuation

Issues

• Covers: False-positives when comparing provenant with scancode-toolkit output #1391

Scope and exclusions

• Included: bounded action-attribution cues with contact evidence, plus contactless handling only in already-bounded author sections
• Explicit exclusions: broad prose attribution without contact or structural author context

How to verify

• Scan Perl 5.38.4 with --copyright and compare against the parent: author records increase from 2,665 to 2,676, recovering 11 net explicit credits while splitting one chained README attribution into precise spans; copyright and holder records are unchanged.
• Scan InfoZIP 3.0 with the same flags and observe identical author, copyright, and holder records to the parent.
• Check that both Configured by root@localhost and aligned Configured by : root@localhost produce no author.

Intentional differences from Python

• ScanCode output is used as evidence, not a correctness target. This layer accepts bounded action vocabulary when a contact or author-section boundary proves intent, including valid credits that ScanCode misses.

Follow-up work

• Created or intentionally deferred: contactless narrative attribution remains a separate benchmark-gated layer.

Expected-output fixture changes

• Files changed: none
• Why the new expected output is correct: unit and scanner contracts cover the new behavior without changing checked-in golden output.

[mstykow]

@greptileai review