Add random seed for running the non-deterministic gs_power_combo()?

[nanxstats]

Running gs_power_combo() with different random seeds:

withr::with_seed(41, gs_power_combo() |> summary() |> as_gt())
withr::with_seed(42, gs_power_combo() |> summary() |> as_gt())
withr::with_seed(46, gs_power_combo() |> summary() |> as_gt())

will give different efficacy for null hypothesis at the last analysis: 0.1955, 0.1956, 0.1957.

If it's ok to have slight variations due to the underlying algorithm, do we want to add a set.seed() to the relevant code examples since the results are not deterministic? Or should this result be much more deterministic theoretically?

For this reason, a with_seed() function had to be used to keep results identical in as_gt() snapshot tests in #337.