Clarify or add default handling for binary predictor prevalence

[GForb]

Summary

When predictor_type = "binary", the wrappers currently require binary_predictor_prevalence to be supplied. If it is omitted, the call eventually fails in generate_predictors() with an error that predictor_prop must be provided.

This behavior is valid but the issue is that the package contract is not fully settled: should binary predictor prevalence be mandatory, or should the wrappers provide a sensible default?

Current behavior

• simulate_binary(), simulate_continuous(), and simulate_survival() accept binary_predictor_prevalence = NULL.
• For binary predictors, generate_predictors() errors when the prevalence is missing.
• As a result, the public interface currently suggests the argument is optional, while the implementation effectively requires it for binary predictors.

Decision needed

Choose one of these directions:

1. Keep prevalence required for binary predictors.
   Then validate early in the wrappers and give a clear user-facing error.

2. Add a default prevalence.
   Then document the default clearly and assess how sensitive estimated sample size is to that choice.

If we add a default

The implementation should include:

• one explicit default prevalence value
• documentation explaining why that default was chosen
• tests covering omitted prevalence for all three wrappers with binary predictors
• a brief sensitivity check or rationale showing the default does not create misleading sample-size recommendations

Goal

Make the interface consistent: either binary predictor prevalence is required and validated clearly, or it has a documented default with justification.