pintervals 1.3.3

Vignette rewritten to match the updated manuscript: the worked example now predicts relative Democratic vote share in the 2024 election from county_turnout_panel (random forest via ranger), adds a summary table of method assumptions and coverage guarantees, and a new appendix evaluating the example with interval_score() and interval_width(). ranger added to Suggests.
Added importFrom("stats", "lm.fit") to the NAMESPACE
Updated maintainer email address

pintervals 1.3.2

county_turnout_panel gains 16 columns (now 38 total):
- total_votes and cvap — the turnout numerator and denominator, so turnout and every share is reproducible from the shipped data.
- votes_dem, votes_rep, votes_other and the matching dem_share, rep_share, other_share — presidential votes and vote shares by party (Libertarian/Green pooled into "other" for a consistent panel).
- eight per-election aggregates, constant within election_year: turnout_county_mean / turnout_electorate and {dem,rep,other}_share_county_mean / _electorate — the unweighted county-wise mean and the pooled electorate-wide ("true") turnout and vote shares for each election, computed over the full electorate.

county_turnout_panel now excludes the 25 micro-county-years whose recorded presidential ballots exceeded the noisy 5-year ACS CVAP estimate (turnout > 1), leaving 15,543 rows. Rows with an unknown (NA) turnout are still retained. The filtering is performed by the build pipeline in data-raw/.

New dataset county_turnout_panel: a tidy five-wave county-level panel of U.S. presidential-election turnout and ACS demographic context for the 2008, 2012, 2016, 2020 and 2024 elections (15,568 rows). Turnout is votes cast divided by Citizen Voting-Age Population (CVAP). All demographic covariates are re-derived directly from the U.S. Census Bureau API (not IPUMS NHGIS, to comply with NHGIS's non-redistribution terms) and match the vintage of each election. Ships with a full codebook at system.file("extdata", "county_turnout_panel_codebook.md", package = "pintervals") and a reproducible build pipeline in data-raw/.
county_turnout (the original single-election 2016 cross-section) is superseded by county_turnout_panel but retained for backward compatibility; it is still used by the package vignette and examples. Its data are unchanged.
Documentation fix: ?county_turnout previously stated that turnout was "total votes cast divided by total population". It is in fact votes divided by the Citizen Voting-Age Population (CVAP); the help page has been corrected. county_turnout_panel$turnout uses this same definition and reproduces the 2016 county_turnout values for 3,103 of the 3,107 shared counties. The Details section of ?county_turnout (the predicted_turnout reproduction code) also now renders — it was previously dropped because of an unescaped % in the Rd source.