Extend discussion of disagreement about replication success (#4)

[LukasWallrich]

Closes #4.

Adds a paragraph to the Hidden Moderator Account section of the Discussion chapter drawing on Sarisoy (2025), Why We Disagree About the Success of Replications (J. Gen. Philos. Sci.).

What it adds

• Names the underdetermination problem explicitly: with limited experimental control and a vast space of possible moderators, experts often cannot tell a genuine replication failure from a violated ceteris paribus assumption (a version of Collins' experimenter's regress, attributed through Sarisoy).
• Reframes such disagreement as legitimate normative judgement rather than poor research practice.
• Introduces Sarisoy's constructive remedy — transparency about a replication's intended epistemic function (testing reliability/stability, a specific validity threat, or generalisation) — and links it to the chapter's existing pre-specification guidance (@sec-success-criteria).

Why here: the section already gestures at irreducible uncertainty ("a single replication is never entirely conclusive"); this anchors that point in the literature and adds the epistemic-function framing, which was not previously covered.

Notes

• New references.bib entry Sarisoy2025; DOI 10.1007/s10838-024-09709-1 verified against Crossref (vol 56(3), 307–324). Inserted alphabetically — trivially mergeable with the in-flight DOI-cleanup PR Verify DOIs across references.bib + clean up 3 entries (#25) #34.
• Rendered discussion.html to confirm the citation resolves inline and in the reference list; render artifacts under docs/ are left to CI, consistent with other content PRs.