Validate init.hiv.prev length matches the race flag (#59)

[smjenness]

Quick fix for #59 (open since 2025-01-09).

What was wrong

Under race = FALSE, the user expected to pass a single overall-population HIV prevalence (length 1), but the docstring said init.hiv.prev was a vector of size 3. The previous active validation was the commented-out

#if (length(init.hiv.prev) != 3) { stop("...vector of size three") }

so anything passed silently. Behavior:

Input	Old behavior	New behavior
length 1 + race = FALSE	Worked (user's request)	Works, now documented
length 3 + race = FALSE	Worked (extra elements ignored)	Works, preserves backward compat
length 3 + race = TRUE	Worked	Works
length 1 + race = TRUE	Cryptic "vector size cannot be NA/NaN" from downstream	Clean error: "init.hiv.prev must have length 3 ..."
length 2 + race = TRUE	Same cryptic failure	Clean error

What changed

• R/EpiStats.R: @param init.hiv.prev doc spells out the length contract per race value. Replaced the commented-out length check with an active one that enforces length(init.hiv.prev) == length(race.level) when race = TRUE, and keeps length >= 1 permissive under race = FALSE (extra elements ignored, with a comment noting that the downstream sampler reads only init.hiv.prev[1]).
• tests/testthat/test-init-hiv-prev.R (new): 5 test blocks covering each combination — valid length-1 + race=FALSE, valid length-3 + race=TRUE, clean errors for length-1 + race=TRUE and length-2 + race=TRUE, out-of-range value still rejected with the original message.

Backward-compat note

The inst/validation/ snapshot harness's atlanta_no_race scenario uses init.hiv.prev = c(0.33, 0.137, 0.084) with race = FALSE. That combination still works under the new validation (length >= 1 is permissive when race = FALSE), and the snapshot match is preserved 3/3.

Test plan

• Backward-compat snapshot 3/3 (atlanta_default, national_no_geog, atlanta_no_race)
• New tests pass (5/5)
• Full testthat suite passes locally
• R CMD check 0/0/0

Closes #59.