fix: chalk boolean style modifiers — <Text dimColor> no longer renders [object Object] (#5039)

[proggeramlug]

Fixes #5039 — ink <Text dimColor> rendered [object Object] instead of its text.

chalk 5's style modifiers are getter accessors on a function prototype chain (Object.defineProperties(createChalk.prototype, styles), reached via Object.setPrototypeOf(chalk, createChalk.prototype) where chalk is itself a function). Four independent gaps conspired so chalk.dim never resolved; the issue's "boolean modifier vs string color" framing turned out to be a red herring — color="cyan" only looked fine because broken chalk made ink fall back to the unstyled string.

The four fixes

1. runtime — closure/dynamic_props.rs: closure_get_dynamic_prop's static-prototype walk only consulted data properties. Both arms (closure proto and plain-object proto) now resolve accessor descriptors, invoking the getter with the original receiver (clone_closure_rebind_this), so chalk's Object.defineProperty(this, styleName, {value: builder}) self-cache lands on the instance instead of throwing Cannot redefine property: dim against the non-configurable accessor on the shared proto.

2. codegen — lower_call/property_get.rs: Annex B HTML-wrapper names (bold, link, anchor, big, …) on any-typed receivers were force-routed to String.prototype, so chalk.bold(s) coerced the chalk closure to its source text and returned <b>(...strings) => strings.join(' ')</b>. Dropped them from the force list; an Any receiver that really is a string still resolves through the jsval.is_string() arm of js_native_call_method (covered in the gap test).

3. resolver — commands/compile/resolve.rs: Node subpath imports (#-prefixed specifiers via the package's own package.json "imports" map) were unsupported, so chalk's import ansiStyles from '#ansi-styles' (vendored dep) silently resolved to nothing and every style table came up empty. Reuses the exports-conditions resolver with node ranked above default (chalk's #supports-color ships a browser build as default). Per Node's package-scope rule only the nearest package.json is consulted.

4. HIR — lower_patterns.rs: unescape_template didn't decode \xHH / \uHHHH / \u{…} (nor \b \f \v \0, line continuations), so ansi-styles' `\^[[${code}m` template literals produced 6 literal chars instead of ESC. Surrogate pairs combine; a lone surrogate becomes U+FFFD (WTF-8 stays a categorical gap).

Validation

• New test-files/test_gap_chalk_proto_styles_5039.ts — byte-identical vs node --experimental-strip-types (chalk's exact proto shape incl. the defineProperty self-cache, nested chalk.dim.bold, Annex B-on-string behavior, template escapes).
• Real chalk 5 compiled via perry.compilePackages produces Node-identical ANSI output for dim/bold/cyan.
• The issue's ink repro (main now includes feat: native yoga-layout via taffy + JSON module imports — ink #348 end-to-end #5038) renders all lines correctly on this branch:

  ^[[2mA dimColor plain ascii^[[22m
  B plain text
  ^[[36mC colored ascii^[[39m
  ^[[1mD bold ascii^[[22m
  

  (cyan now emits real escapes too — previously it silently rendered unstyled.)
• 3 new resolver unit tests for the imports map (exact, conditional node/default, unmapped).
• cargo test --release for perry-hir / perry-codegen / perry-runtime / perry: green (issue_4908_subclass_native_member_base failed once under full-suite load, passes in isolation — known load artifact).
• Full parity sweep (886-file corpus): 716 pass / 47 parity-fail / 0 compile-fail (93.8%). All 47 fails verified pre-existing: re-ran the exact subset on a baseline build of a46ca1dd6 (pre-fix main) — 46 fail identically there, and the 47th (test_issue_2131_net_lifecycle_edge) is the documented load flake (passes 100% idle on this branch).

Deferred (pre-existing, observed while validating)

• Object.entries(o) includes non-enumerable defineProperty slots (Object.keys is correct): Object.defineProperty(o,'hidden',{value:1}); o.shown=2 → Perry [["hidden",1],["shown",2]], Node [["shown",2]]. Observed as ansi-styles enumerating 55 entries vs Node's 45 — harmless for chalk (extra unused getters).
• Perry's JSON.stringify emits the 4-hex-digit unicode escapes for U+000C/U+0008 instead of the \f/\b short forms — filed as JSON.stringify emits \u000c/\u0008 instead of the \f/\b short forms #5047. (The Object.entries finding above is Object.entries/values include non-enumerable defineProperty slots (Object.keys is correct) #5046.)

Rebased onto main after #5038/#5040/#5043/#5041 landed; this PR is v0.5.1163.