parser: friendly error for fld used as binding name

[danieljohnmorris]

Summary

fld=5 used to silently parse as a 0-arg call to the fold builtin and surface as ILO-T006 arity mismatch: 'fld' expects 3 or 4 args, got 0. Personas reach for fld as a natural variable (field/fold/folder) and pay the retry tax on a misleading error.

This adds fld to the existing reserved-word friendly-error path that already covers cnt/brk/var/let/if/etc. (the work from 8928635). Same family, same shape, same ILO-P011 code.

Repro

Before:

$ ilo run -e 'f>n;fld=5;p fld'
ILO-T006 arity mismatch: 'fld' expects 3 or 4 args, got 0
ILO-T005 undefined function 'p' (called with 1 args)

After:

$ ilo run -e 'f>n;fld=5;p fld'
ILO-P011 `fld` is reserved for the fold builtin and cannot be used as an identifier
hint: pick a different name like `field` or `folder`

What's in the diff

• parser: friendly error for fld used as binding name — adds the guard in parse_decl and parse_stmt, mirroring the existing cnt/brk shape so the friendly error fires uniformly at top level and inside function bodies.
• tests: regression coverage — tests/regression_fld_reserved.rs pins the ILO-P011 path under tree/vm/cranelift for both body-level and inside-loop bindings, plus a sanity test that the fld builtin itself still works. A parallel one-liner added to regression_friendly_errors.rs keeps the friendly-error suite the canonical reserved-word audit. examples/fld-reserved-rename.ilo shows the renamed-variable shape an agent should reach for after hitting the error.

Test plan

• cargo test --release --features cranelift green (all suites)
• cargo fmt --check clean
• cargo clippy --release --features cranelift --tests clean
• Manual repro confirmed before/after across tree/vm/cranelift

Follow-ups

None.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!