Make type of Assign more specific with Variable inductive

[keyboardDrummer-bot]

Summary

Implements the refactoring requested in #21: makes the type of Assign targets more specific by introducing a Variable inductive type.

Changes

Core type change (Laurel.lean)

• Introduced Variable inductive inside the mutual block:

  inductive Variable where
    | Local (name : Identifier)
    | Field (target : AstNode StmtExpr) (fieldName : Identifier)

• Changed Assign from List (AstNode StmtExpr) to List (AstNode Variable)
• Replaced StmtExpr.Identifier with StmtExpr.Variable (.Local name)
• Removed StmtExpr.FieldSelect, replaced with StmtExpr.Variable (.Field target fieldName)
• Added VariableMd abbreviation

Grammar (LaurelGrammar.st)

• Added multiAssign operator for multi-target assignment syntax

Updated passes

• MapStmtExpr.lean — updated traversal for new constructors
• Resolution.lean — added resolveVariable in mutual block with resolveStmtExpr
• HeapParameterization.lean — updated field detection and heap variable references
• LaurelToCoreTranslator.lean — updated pattern matching on targets
• LiftImperativeExpressions.lean — updated assign target handling
• FilterPrelude.lean, CoreGroupingAndOrdering.lean, InferHoleTypes.lean, ConstrainedTypeElim.lean, EliminateHoles.lean, EliminateValueReturns.lean, ModifiesClauses.lean, TypeHierarchy.lean, LaurelTypes.lean
• Grammar translators (both directions)
• Python-to-Laurel translator and specs

Notes

• Multi-target assignment in the grammar uses identifier targets only (as specified in the issue)
• The assign comment documents that multiple targets only work with identifier targets and a call as RHS

Closes #21

[keyboardDrummer-bot]

I investigated the CI failure. The only failing check is "Run internal benchmarks of Strata" which requires AWS credentials (secrets.AWS_BENCHMARK_ACCOUNT) to trigger a CodeBuild job. The job failed in ~1 minute (started 10:30:15, completed 10:31:23), which is consistent with a credential/permissions failure at the "Configure AWS credentials" step rather than an actual code issue.

All other 13 checks passed (Lean build+test, Python tests on 3.11–3.14, CBMC tests, lint, documentation, etc.).

Locally, lake build, lake test, and lake lint all pass successfully. This failure is not caused by the code changes in this PR — it's an infrastructure issue with the benchmark job's AWS access in this repository.