Add author correction and DCP follow-up analysis

[GoGoKo699]

This pull request updates the repository with:

• an author-maintained correction of the 2022 DCP challenge analysis;
• the exact 23/32 versus 25/32 counterexample;
• the one-sample optimality theorem;
• exact honest success probabilities and corrected Figure 5 values;
• corrected statistical uncertainty;
• a separate phase-twirled heralded DCP witness;
• historical context and related maximum-confidence literature;
• explicit GPT-5.6 Pro assistance disclosure;
• reproducible code, tests, results, PDF, and validation records.

The original six Python scripts remain unchanged, and the original repository state is preserved at the immutable paper-2022-original release.

The correction is author-maintained, is not an APS Erratum, and does not alter the published journal article.