Fix conditional beta calculations and clarify gs_cp_npe2() behavior

[LittleBeannie]

Summary

This PR corrects the conditional beta probability calculation in gs_cp_npe2() and clarifies the function's behavior when the next analysis is the target analysis (j = i + 1).

Related

Closes #660
Closes #661

Changes

• Correct the upper integration limits used to calculate prob_beta when j > i + 1.
• Use the efficacy bounds b instead of the futility bounds a as the upper limits for intermediate analyses.
• Expand the return-value documentation for prob_alpha, prob_alpha_plus, and prob_beta.
• Explicitly distinguish the formulas for j = i + 1 and j > i + 1.
• Add a regression test comparing gs_cp_npe2() with gs_cp_npe1() when j = i + 1.
• Add a regression test based on issue Update gs_cp_npe2 beta calculations #661, verifying that the probability of crossing an efficacy boundary plus the probability of not crossing through the final analysis sums to 1.

Results

For the example reported in #661, prob_beta[2] is corrected from 0 to approximately 0.4437:

sum(result$prob_alpha) + result$prob_beta[2]
# 1