Changes in version 0.95.1                        

  - CRAN-safe multiprecision default: builds now use Boost cpp_dec_float
    from BH with DW_USE_MPFR=0; local source builds can opt in to
    MPFR/GMP with DW_USE_MPFR=1.
  - Added rootWishartHD_mpfr_enabled() and runtime warnings when the
    deprecated MPFR-specific interface is requested but MPFR was not
    compiled.
  - Renamed the main user-facing workhorse functions to remove
    experimental from the public API: doubleWishart_log(),
    doubleWishart_scaled(), singleWishart_log(), and
    singleWishart_cdf().
  - Renamed the global runtime option to
    options(rootWishartHD.force_multiprecision = TRUE).
  - Activated tinytest via tests/tinytest.R and kept long numerical
    sweeps under inst/validation/ for manual release validation.
  - Reduced vignette Monte Carlo examples and kept multi-core benchmark
    code in non-evaluated chunks.
  - Hardened theta-scale validation, large-s routing,
    p-value/critical-value fallbacks, and single-Wishart input
    validation.