fix(test/docs): unblock CI on main — known_failures + styling.md fences

origin/main has been red on the Tests workflow since v0.5.310/.314/.316
landed. Two unrelated unblockers:

1) test-parity/known_failures.json — add `test_inline_uint8array_param`
   (added v0.5.314, #169 fix) and `test_issue_167_loop_alloca_stack_eat`
   (added v0.5.316, #167 fix). Both compile-fail on macOS-14 CI runners,
   matching the existing test_gap_buffer_ops family pattern. Status: bug
   so they remain visible on the triage list.

2) docs/src/ui/styling.md — two illustrative `typescript` fences (lines
   49 + 70 from the v0.5.310 inline-style doc work) showed object-literal
   fragments that aren't standalone programs. Tag them `typescript,no-test`
   to opt them out of the doc-test compile pass. The doc-test lint added
   in the v0.5.304 audit was rejecting them because they weren't tagged
   and weren't `{{#include}}` directives.

This unblocks the parity + macOS doc-tests gates on every open PR.

Author: proggeramlug

docs/src/ui/styling.md

@@ -46,7 +46,7 @@ for the full file.
 
 Color props accept four interchangeable shapes:
 
-```typescript
+```typescript,no-test
 backgroundColor: "#3B82F6"                                   // hex 6/8
 backgroundColor: "#3B82F6FF"                                 // hex with alpha
 backgroundColor: "blue"                                      // named color
@@ -67,7 +67,7 @@ so `backgroundColor: someStringVar` works the same as the literal form.
 
 A single number applies to all four sides; an object picks per-side:
 
-```typescript
+```typescript,no-test
 padding: 12                                       // all four sides 12
 padding: { top: 8, right: 16, bottom: 8, left: 16 }  // per-side
 ```

test-parity/known_failures.json

@@ -62,5 +62,13 @@
   "test_stress_promises": {
     "status": "bug",
     "reason": "Multiple distinct Promise issues remaining after v0.5.283 fix (which addressed (a) microtask FIFO ordering — TASK_QUEUE was a Vec with .pop() → LIFO; switched to VecDeque + pop_front, and (b) `.then(h)` / `.catch(h)` not catching throws inside h — microtask runner now wraps closure call in setjmp): (1) `new Promise<T>((resolve, reject) => ...)` constructor body not running (lines `constructor resolve: 99` / `constructor reject: constructor-err` missing from output); (2) `Promise.any` produces no output (any/any-all-rejected/any-errors-length lines missing); (3) `.then(p => Promise.resolve(...))` doesn't unwrap inner Promise (logs `Promise { <pending> }` instead of resolved value); (4) `.finally(() => ...).then(v => ...)` value-passing wrong (Perry: `finally value: 0`, Node: `finally value: done`)."
+  },
+  "test_inline_uint8array_param": {
+    "status": "bug",
+    "reason": "Regression test added by v0.5.314 (#169 fix in perry-transform substitute_locals). The test compiles successfully locally but compile-fails on CI runners — likely an SDK/linker version interaction shared with the test_gap_buffer_ops family. The underlying Uint8Array param substitution fix landed in 976c3eb5; this test documents the case but the CI env exhibits a separate compile-fail mode. Pair-investigate with test_issue_167_loop_alloca_stack_eat."
+  },
+  "test_issue_167_loop_alloca_stack_eat": {
+    "status": "bug",
+    "reason": "Regression test added by v0.5.316 (#167 fix hoisting js_native_call_method args alloca to entry block). Same compile-fail-on-CI / passes-locally pattern as test_inline_uint8array_param — both tests are about Buffer/Uint8Array codepaths and likely share the macOS-14 SDK gap."
   }
 }