Vectorized Bayesian bootstrap

[VisruthSK]

This PR was made with assistance from Codex.

Swaps the Bayesian bootstrap implementation from a loop to matrix ops. Adds a test for BB weights correctness before changing implementation.

Benchmark:

set.seed(0)
N <- 400
K <- 6
BB_n <- 1000
alpha <- 1
lpd_point <- matrix(rnorm(N * K, mean = -1), nrow = N, ncol = K)
BB_weighting <- loo:::dirichlet_rng(BB_n, rep(alpha, N))

pseudobma_bb_loop <- function(lpd_point, BB_weighting) {
  N <- nrow(lpd_point)
  K <- ncol(lpd_point)
  BB_n <- nrow(BB_weighting)
  temp <- matrix(NA, BB_n, K)

  for (bb in 1:BB_n) {
    z_bb <- BB_weighting[bb, ] %*% lpd_point * N
    uwts <- exp(z_bb - max(z_bb))
    temp[bb, ] <- uwts / sum(uwts)
  }

  colMeans(temp)
}

pseudobma_bb_vectorized <- function(lpd_point, BB_weighting) {
  N <- nrow(lpd_point)
  z <- BB_weighting %*% lpd_point * N
  uwts <- exp(z - matrixStats::rowMaxs(z))

  colMeans(uwts / rowSums(uwts))
}

results <- bench::mark(
  loop = pseudobma_bb_loop(lpd_point, BB_weighting),
  vectorized = pseudobma_bb_vectorized(lpd_point, BB_weighting),
  iterations = 1e3
)

summary(results)
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 loop          4.4ms   5.25ms      188.    9.16MB    21.8 
#> 2 vectorized   1.02ms   1.19ms      814.  167.95KB     2.45
summary(results, relative = TRUE)
#> # A tibble: 2 × 6
#>   expression   min median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
#> 1 loop        4.32   4.41      1         55.8     8.91
#> 2 vectorized  1      1         4.33       1       1

^{Created on 2026-06-06 with reprex v2.1.1}

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.69%. Comparing base (286263d) to head (230331c).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #368      +/-   ##
==========================================
- Coverage   92.70%   92.69%   -0.01%     
==========================================
  Files          31       31              
  Lines        3029     3025       -4     
==========================================
- Hits         2808     2804       -4     
  Misses        221      221              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 230331c is merged into master:

• ✔️loo_function: 1.77s -> 1.77s [-0.24%, +0.54%]
• ❗🐌loo_matrix: 1.29s -> 1.31s [+0.54%, +1.3%]
  Further explanation regarding interpretation and methodology can be found in the documentation.