fix(parser): enforce trailing semicolon on field_decl

Previously parseFieldDecl discarded the result of
`expect(p, Semicolon)` via `let _ = ...`, silently accepting a
missing trailing `;` on a region field. That contradicted EBNF
grammar.ebnf §field_decl which requires the semicolon.

Fix: propagate the Semicolon error. L1 test suite gains a case
locking in the stricter behaviour. All other suites still pass:

  - ParserTests: 88/88 (unchanged)
  - E2E driver:   4/4 clean, 2 honest-skip for L11/L12
  - L1-L3:        17/17 (was 16; L1 grew from 7 to 8)

Scope sweep confirmed no regression risk: grep for field-like
lines without `;` surfaced 12 hits, all function parameters
(comma-separated, not field_decls). No example / test fixture
relied on the old looseness.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: hyperpolymath

src/parser/Parser.res

@@ -1331,8 +1331,14 @@ let parseFieldDecl = (p: parserState): result<Ast.located<Ast.fieldDecl>> => {
           | Error(_) => ()
           }
         }
-        let _ = expect(p, Semicolon)
-        Ok({node: {name, fieldType: ty, constraints}, loc: startLoc})
+        // EBNF grammar.ebnf §field_decl requires the trailing semicolon.
+        // Older fixtures relied on a silent `let _ = expect(...)` that
+        // discarded the error; enforce it now so parse-time L1 coverage
+        // matches the spec.
+        switch expect(p, Semicolon) {
+        | Ok() => Ok({node: {name, fieldType: ty, constraints}, loc: startLoc})
+        | Error(e) => Error(e)
+        }
       | Error(e) => Error(e)
       }
     | Error(e) => Error(e)

tests/levels/L1.mjs

@@ -66,11 +66,9 @@ expectOk(
 
 // ── Negative: canonical syntax errors ────────────────────────────────
 
-// Finding 2026-04-18: parser is lenient about the semicolon between
-// the last field and `align`. Grammar EBNF requires `;` after every
-// field_decl, so this is a parser looseness worth flagging in a
-// future parser-tightness pass. Not exercised here to keep L1 tests
-// aligned with current parser behaviour rather than the EBNF spec.
+// Parser now enforces the trailing `;` on every field_decl per
+// grammar.ebnf §field_decl. A missing `;` between the last field and
+// the region's `align` line is rejected rather than silently accepted.
 
 expectErr(
   `region Bad { n: u32;
@@ -79,6 +77,14 @@ expectErr(
   "align with missing integer literal",
 );
 
+expectErr(
+  `region Bad {
+    n: u32
+    align 4;
+  }`,
+  "field decl missing trailing semicolon",
+);
+
 expectErr(
   `region Bad {
     n: u32;