Fold a field projection through a foreign-header scope IIFE - #161

A ForeignImport with a header lowers to a chunk [ForeignSourceStat fh, return exportsTable]. Projecting a field off it lowers to (function() ...; return exports end)().field — a shape neither the IR-level reduceObjectProp nor the Lua-level reduceTableDefinitionAccessor sees through, since the projected value is a function call, not a table constructor, at either level.

Summary

• Added foldFieldProjectionThroughScopeCall, a Lua-AST rewrite rule that projects the field inside a no-arg immediately-invoked function whose last statement is a return, rather than after the call: (function() …; return e end)().foo → (function() …; return e.foo end)(). Sound because the projection happens within the same activation regardless of whether it's before or after the call returns, so no side effect crosses the boundary.

• The rewritten projection is immediately re-optimized, so it composes with reduceTableDefinitionAccessor in the same pass. For the Data.Eq.eqChar case from the issue, this fully collapses:

  ((function()
    local refEq = function(r1) return function(r2) return r1 == r2 end end
    return { eqCharImpl = refEq }
  end)()).eqCharImpl

  down to:

  (function()
    local refEq = function(r1) return function(r2) return r1 == r2 end end
    return refEq
  end)()

• Wired into rewriteRulesInOrder alongside the existing rules.

• Added unit tests covering full reduction, partial fold (when the returned value isn't an unambiguous table constructor), and decline (callee not a no-arg IIFE).

Checklist

• Added a changelog.d/ fragment for any user-facing change (scriv create
  in the dev shell), or this change ships nothing releasable (CI, docs, or an
  internal refactor).
• In the dev shell (nix develop), fourmolu -i lib/ exe/ test/ and
  hlint lib/ exe/ test/ are clean.
• In the dev shell, cabal test all passes; structural goldens were
  re-accepted on purpose if codegen moved (PSLUA_GOLDEN_ACCEPT=1), and
  eval/golden.txt still holds.