Add mini meta-analysis guidance for mixed original results (#64)

[LukasWallrich]

Adds one short paragraph to the "Uncertainty" section of choosing_study.qmd, directly after the power-deflation paragraph, covering the case where the original paper reports a mix of significant and non-significant results.

The paragraph makes three points:

• An internal or "mini" meta-analysis of the original studies can pool their effect estimates into one estimate with a confidence interval, which is generally more informative than counting significant studies, provided the studies are comparable enough and the analysis accounts for precision and heterogeneity (Goh, Hall & Rosenthal, 2016).
• Differences between the studies (samples, operationalizations, settings) can generate hypotheses about boundary conditions for a replication to test, while noting that such differences are post hoc and often confounded.
• A pooled estimate from a single lab's own studies can itself be distorted by selective reporting, so the caution in the preceding paragraph still applies (Vosgerau, Simonsohn, Nelson & Simmons, 2019).

GohEtAl2016 already existed in references.bib. VosgerauEtAl2019 is new, with Crossref-verified metadata (doi 10.1037/xge0000663); the % in the title is escaped for BibTeX.

Chapter renders without citation or cross-reference warnings.

Closes #64