feat: simulate_chain_stats() returns both size and length statistics

[Copilot]

simulate_chain_stats() only returned the statistic used as the stopping criterion, making it impossible to limit by one stat (e.g., generation length) while observing the other (e.g., outbreak size).

Changes

Core API change: <epichains_summary> is now a data frame

• <epichains_summary> changes from a single-column numeric vector to a data.frame with size and length columns
• statistic remains the stopping criterion; only that stat is censored at stat_threshold (the complementary stat reflects its actual value when the chain stopped)

# Limit by generation length, observe outbreak sizes
simulate_chain_stats(
  n_chains = 10, offspring_dist = rpois,
  statistic = "length", stat_threshold = 5, lambda = 1.5
)
#> `epichains_summary` object
#>
#>    size length
#> 1    14      8   # length hit threshold → Inf; size is actual count
#> 2     2      2
#> 3    44    Inf   # censored at threshold
#> ...
#>  Stopping criterion: length (threshold: 5)

Updated consumers

• format.epichains_summary() — displays both stats with stopping criterion and threshold
• summary.epichains_summary() — returns max_size/min_size/max_length/min_length (replaces max_stat/min_stat)
• summary.epichains() — computes and returns both stats
• likelihood() — extracts the relevant column from the data frame; uses stat_threshold only when the requested statistic matches the stopping criterion (otherwise Inf, since the non-stopping stat is never censored)
• .offspring_ll() (stat_likelihoods.R) — extracts statistic column before passing to ecdf()
• rborel() (borel.R) — extracts size column from the new data frame

Tests & snapshots

• Tests updated throughout to use $size/$length, nrow(), and the renamed summary fields
• Snapshots regenerated to reflect the new data frame print format