Changelog of @hackage/unwitch 3.0.0

Change log for unwitch project

Int32.toInt is now total on GHC (Int32 -> Int instead of Int32 -> Maybe Int). GHC's Int# is always at least 32 bits wide (WORD_SIZE_IN_BITS), so every Int32 value fits without loss. Code pattern matching on Just/Nothing will need updating.
On non-GHC compilers, Int32.toInt retains the Maybe Int return type since the Haskell Report only guarantees Int has 30 bits.
CInt.toInt now delegates to Int32.toInt instead of using fromIntegral.

Add MicroHs CI support — all exposed modules are compiled with mhs
Guard GHC-specific code behind CPP: unboxed types, GHC.Exts imports, MagicHash/UnboxedSums/UnboxedTuples extensions, and ghc-bignum dependency
Guard CInt module and all toCInt functions behind GLASGOW_HASKELL (MicroHs defines CInt as a newtype over Int, not Int32)
Replace GHC.Float.double2Float/float2Double with portable realToFrac
Replace Data.Text.all with Prelude.all for MicroHs compatibility
CI module list is now extracted from the cabal file automatically

New module Unwitch.Convert.CInt — conversions from CInt (Foreign.C.Types) to all supported numeric types (Int, Int8–64, Word, Word8–64, Integer, Natural, Float, Double)
Add toCInt to all 14 existing Convert modules (Int, Int8, Int16, Int32, Int64, Integer, Word, Word8, Word16, Word32, Word64, Natural, Float, Double)
Add CIntSpec test module with 12 tests covering total and fallible CInt conversions

Add Haddock documentation to partial conversion functions across all Convert modules — docs focus on non-obvious information only: float/double precision limits, via-Integer failure modes (NaN, infinity, fractional values), Underflow for negative-to-Natural, and Latin-1 encoding constraints
Remove redundant docs that merely restated the type signature

Wire up 9 missing test specs that existed but were not included in the test runner
Add module-level haddock to all Convert modules, Errors, and Constant
Add haddock to Overflows type and constructors
Expand property tests with 20+ new properties covering Int/Word same-width conversions, Float/Double range checks, cross-sign narrowing, Ratio round-trips, and Fixed round-trips

Add haddock section headers separating boxed and unboxed conversions
Each module with unboxed functions now has "Conversions" and "Unboxed conversions" sections with documentation explaining zero-allocation failure handling and required GHC extensions.

Implement some of Double. Write description