fix(hir): error on widget method-chain modifier syntax (closes #195)

[proggeramlug]

Summary

• Option B chosen: reject method-chain modifier syntax at compile time with a clear diagnostic. Option A (fold chain into constructor options-object) was deliberately deferred per the issue guidance — it requires a multi-hundred-line HIR pre-pass.
• Adds is_widget_modifier_name helper + a ~28-line detection block inside lower_expr's ast::Expr::Member handler.
• Adds 6 regression tests in crates/perry-hir/tests/widget_modifier_diagnostic.rs.
• Minimal docs update: rewrites the "Text Modifiers" example in docs/src/widgets/components.md to use the options-object form and adds a one-line note. The larger "Modifiers" section (lines 116–175) examples still show the chained form; updating all of them would exceed ~30 lines — left for a follow-up.

Modifiers recognised

font, fontWeight, weight, foregroundColor, color, foreground, padding, cornerRadius, background, backgroundColor, opacity, lineLimit, frame, minimumScaleFactor, containerBackground, maxWidth, url, bold, italic, underline, fontSize, strikethrough, multilineTextAlignment, lineSpacing

Example error output

error: modifier 'font' must be passed as an option-object on the widget constructor; use: Text("...", { font: ... })

Chained calls fail on the first modifier via natural recursion — processing .color(...) tries to lower the inner Text(...).font(...), which fails, and the error propagates.

Scope / known limitation

The check lives in lower_expr, which is the path for standalone perry/ui code. Method-chain modifiers inside Widget({...}) render bodies (from perry/widget) travel a separate AST-only path (parse_widget_node) that never calls lower_expr, so they still silently drop rather than error there. Fixing that path requires changing parse_render_body_stmts to return Result — deferred as a follow-up.

Test plan

• cargo test -p perry-hir --test widget_modifier_diagnostic — 6 new tests pass
• cargo build --release -p perry-runtime -p perry-stdlib -p perry — clean build
• Full workspace test suite — 0 failures

https://claude.ai/code/session_01Q6AhxeUKokdecgMWrqyzkB

---

Generated by Claude Code