Add effect/heterogeneity and normalized reports

[FBartos]

Overview

This PR extends phackR so simulations can be generated under a nonzero true effect and between-study heterogeneity, and makes the resulting analyses report standardized initial/final summaries in a consistent format.

Main Changes

• Added effect and heterogeneity arguments to all exported sim.* functions.
• Defined one iteration as one study with a study-level draw theta_i ~ Normal(effect, heterogeneity).
• Implemented family-specific effect scales:
  • group-comparison simulators: standardized mean-difference scale
  • regression / correlation simulators: Fisher-z scale
  • covariate simulator: group-effect scale
• Added a normalized reporting block to all exported simulators:
  • effect.initial, effect.final
  • se.initial, se.final
  • n.initial, n.final
  • stat.initial, stat.final
  • p.initial, p.final
  • method.initial, method.final
• Preserved legacy output columns such as ps.*, r2s.*, ds.*, and eta2s.* for compatibility.

Implementation Notes

• Added shared helpers for study-level effect draws, Fisher-z handling, standardized mean-difference reporting, association reporting, ANCOVA/group-effect reporting, and normalized result assembly.
• Extended the previously deferred simulators, including cutoffs, variable transformations, favorable imputation, outlier exclusion, composite scores, covariates, and the regression branch of selective reporting of IVs.
• Updated the regression branch of sim.multIVhack() so alternative is respected for one-sided tests.
• Fixed one-sided direction handling in the group-comparison family so alternative = "greater" now means a positive effect in the intended direction: treatment/second group > control/first group.
• Fixed the related sign inconsistency in legacy ds.* outputs.

Shiny App Updates

• Added effect and heterogeneity controls to the Shiny UI for all relevant tabs.
• Updated the app wording so rates are described as significant-result / rejection rates outside the null-only regime.
• Clarified the one-sided option in the UI and docs as testing a positive effect: treatment > control.
• Fixed the Selective Reporting IV tab to call sim.multIVhack() instead of sim.multDVhack().
• Updated the Shiny startplot generation script and regenerated startplots.rds.

Documentation and Tests

• Updated the vignette and Shiny markdown documentation to describe the new arguments, effect scales, reporting fields, and one-sided semantics.
• Regenerated roxygen documentation and NAMESPACE.
• Expanded the test suite to cover:
  • normalized reporting columns
  • non-null behavior
  • heterogeneity reproducibility
  • one-sided direction behavior

Verification

• testthat::test_file('phackR/tests/testthat/test-simfunctions.R')
• knitr::knit('phackR/vignettes/phackR_vignette.Rmd', ...)
• direct smoke simulations for the extended simulator families
• Shiny app source check after UI/server updates