Optimize CSS selector traversal and fix sibling combinators:

- inline hot-path node and stylesheet traversal accessors
  - correct adjacent and general sibling matching direction
  - preserve inverse sibling operator behavior
  - add selector traversal and sibling regression tests

.agent/plans/eepp_css_selector_optimization_plan.md

@@ -60,6 +60,8 @@ Do not start with a large rewrite. Measure after each phase.
 
 # Phase 1: Inline Tiny Hot-Path Accessors
 
+**Status: Implemented**
+
 ## Motivation
 
 CSS selector matching repeatedly calls small accessors and type checks while walking parents/siblings and matching selector rules. These functions are ideal candidates for header inlining.
@@ -189,6 +191,8 @@ Create an HTML tree containing a normal eepp widget between/inside HTML elements
 
 # Phase 2: Fix Confirmed Sibling Combinator Direction Bug
 
+**Status: Implemented**
+
 ## Motivation
 
 This is a confirmed correctness bug, not intentional behavior. Right-to-left selector matching for CSS combinators should follow CSS semantics:
@@ -1142,4 +1146,3 @@ Inlining is worthwhile but not the main algorithmic win. The main problem is can
 - Keep string fields even after adding hashes to avoid breaking public APIs, debug output, CSS serialization, and tests.
 - Prefer minimal changes with tests after each phase.
 - Do not implement signature caching or sort redesign until instrumentation proves they are still needed.
-