pin down the implementation of bootstrap in anchored case

[hoppanda]

option 1: bootstrap CI for IPD eff measure, then bucher with AgD eff measure
option 2: in the bootstrap of IPD, lump via monte carlo the uncertainty of AgD eff measure [pro of this approach is one can use any ci type from boot R package]

both implemented in v1, so the team can evaluated via simulation in manuscript