….1 (R6/#595)

Three v0.26.0 enum/parser fixes, survey-driven (79% → 85%, 241 → 262/307):

- ENUM-CASE PARAM DEFAULT (illegalstudio#468, declarations.rs): `E $x = E::Case` — an
  enum-case default has the enum object type, but syntactic inference types
  every `::` access as Str. When the declared type is Object(E) and the
  default is a scoped access naming E, accept it (the semantic pass validates
  the case exists). Clears the Badge/Button `$variant expects Object, got Str`
  family (9).
- KEYWORD ENUM-CASE NAMES (body.rs): PHP 8 allows semi-reserved keywords as
  enum-case names (`case Default;`, `case String;`, `case Float;`) — every one
  but `class` (reserved for `Foo::class`). Clears CardVariant/LogicalType/
  HeadAssetMode `Expected case name after 'case'` (10).
- USE CONST/FUNCTION IMPORTS (illegalstudio#473/EC-10/illegalstudio#493, namespace_use.rs): the lexer
  eagerly tokenizes `PHP_INT_MAX` etc., so the use-declaration parser must
  accept those dedicated tokens as import names (`use const PHP_INT_MAX;`).
  Clears RegionEntry `Expected imported name after 'use'` (2).

Deferred (not survey-visible on v0.26.1): illegalstudio#469 keyword-case access, illegalstudio#472 clone.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…opy (R9/#599, illegalstudio#472)

v0.26.1's parser rejected `clone $x` ("Expected ';'"), gapping every PSR-7
immutability wither (Message/Uri/RequestBehavior `$new = clone $this`). Re-ports
illegalstudio#472 (8485e9e) onto the post-EIR-refactor tree:

- PARSER: contextual `clone <expr>` prefix operator (clone is not a reserved
  token — `clone_operand_follows` guards it so a bare identifier spelled "clone"
  in a non-operator position still parses as a name). New ExprKind::Clone(Box<Expr>).
- CHECKER: `clone $x` infers as the type of `$x`.
- EIR: shallow object/array copy lowering (array_clone_shallow, x86 frame-safe).
- The new AST variant is threaded through every ExprKind match (name resolver,
  optimizer passes, magic-constant walker, usage/invalidation analysis, warnings).
  The legacy backend is gone in v0.26.1, so no legacy emit arm is needed.

clone codegen tests pass (round-trip + contextual-keyword). Survey 90% → 91%
(277 → 278/307); the "Expected ';'" gaps (Message ×4, Uri, RequestBehavior) clear.
Also completes R6's deferred illegalstudio#472.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…opy (R9/#599, illegalstudio#472)

v0.26.1's parser rejected `clone $x` ("Expected ';'"), gapping every PSR-7
immutability wither (Message/Uri/RequestBehavior `$new = clone $this`). Re-ports
illegalstudio#472 (8485e9e) onto the post-EIR-refactor tree:

- PARSER: contextual `clone <expr>` prefix operator (clone is not a reserved
  token — `clone_operand_follows` guards it so a bare identifier spelled "clone"
  in a non-operator position still parses as a name). New ExprKind::Clone(Box<Expr>).
- CHECKER: `clone $x` infers as the type of `$x`.
- EIR: shallow object/array copy lowering (array_clone_shallow, x86 frame-safe).
- The new AST variant is threaded through every ExprKind match (name resolver,
  optimizer passes, magic-constant walker, usage/invalidation analysis, warnings).
  The legacy backend is gone in v0.26.1, so no legacy emit arm is needed.

clone codegen tests pass (round-trip + contextual-keyword). Survey 90% → 91%
(277 → 278/307); the "Expected ';'" gaps (Message ×4, Uri, RequestBehavior) clear.
Also completes R6's deferred illegalstudio#472.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>