fix: preserve full 256-bit precision for integer literals

[N0xMare]

Summary

Fixes #40 — Integer literals >2^64 were silently truncated to u64.

Also upgrades typeck const evaluation from u64 to full U256 precision.

---

Commit 1: fix: preserve full 256-bit precision for integer literals

Problem

Lit::Int(u64) in the AST meant the parser accumulated lexer bytes into u128 then cast to u64, destroying precision. Any literal value above 18,446,744,073,709,551,615 was silently wrong — blocking MAX_UINT, realistic token amounts (e.g. 10^21 wei), and address literals.

The lexer already produced correct [u8; 32] big-endian bytes via decimal_to_bytes32(). The parser threw them away.

Fix

• Lit::Int(u64, ...) → Lit::Int([u8; 32], ...) — carry raw bytes through the AST
• Parser — copy lexer bytes directly instead of the lossy u128→u64 accumulation
• IR lowering — values <2^63 use SmallInt(i64) (same path as before), values ≥2^63 use LargeInt(hex_string) which codegen already handles correctly
• Typeck eval_const_expr — overflow guard for values >u64 (superseded by commit 2)
• All other downstream consumers (array index/length extraction) updated

No behavioral change for small values — values 0 to 2^63-1 follow the exact same SmallInt code path.

Tests added (commit 1)

• 3 parser tests: 2^64 preserved, MAX_UINT bytes correct, small value unchanged
• 4 e2e bytecode tests: 2^64, 10^21, 2^256-1, and 42 all produce correct PUSH instructions

---

Commit 2: feat(typeck): upgrade const eval to full U256 precision

Problem

eval_const_expr worked on u64 — constants >2^64 returned 0, arithmetic overflowed at 10^19, and shifts were masked with & 63 instead of % 256.

Fix

• ConstValue.value: u64 → [u8; 32] (big-endian 256-bit)
• eval_const_expr: rewritten to use ruint::aliases::U256 arithmetic internally
  • All arithmetic uses 256-bit wrapping (matches EVM semantics)
  • Shifts check r >= 256 → return zero (correct EVM behavior)
  • Division/modulo use is_zero() guard
  • Lit::Int passes [u8; 32] through directly — no truncation
  • Lit::Hex/Bin properly pads into 32 bytes

Self-contained in typeck — ConstValue is not consumed outside the typeck crate. IR lowering re-evaluates constants from AST expressions independently.

Tests added (commit 2)

• Large literal >u64::MAX: verifies exact bytes
• Const arithmetic (A * A): verifies computed value
• Shift left 128: verifies correct 256-bit result (was zero with u64)
• Small u8 value: verifies no regression

---

All 670+ workspace tests pass.

[netlify]

✅ Deploy Preview for edgelang ready!

Name	Link
🔨 Latest commit	1e4a3d7
🔍 Latest deploy log	https://app.netlify.com/projects/edgelang/deploys/69aec61c059760000857d213
😎 Deploy Preview	https://deploy-preview-55--edgelang.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[brockelmore]

lgtm pending merge of main and resolution of that conflict