perf: drop Mutex from cached_dasherize

[tomasc]

Summary

Removes the Mutex#synchronize wrapper around cached_dasherize's cache write. The operation is deterministic — same input, same output — so a concurrent duplicate compute produces the same value and the overwrite is harmless. Ruby Hash writes are atomic enough at the VM level for this pattern (used widely in Rails internals).

Net effect: no more lock acquisition + no more Mutex#synchronize block allocation on every call. cached_dasherize is called hundreds of times per heavy page render in consumer apps (rijksakademie.nl's MemoryProfiler showed this gem at 1.5 MB allocated per request).

Code diff

-  @cache_mutex = Mutex.new
-  class << self
-    attr_reader :dasherized_cache, :cache_mutex
-  end
+  class << self
+    attr_reader :dasherized_cache
+  end

   def cached_dasherize(string)
-    StimulusHelpers.cache_mutex.synchronize do
-      StimulusHelpers.dasherized_cache[string] ||= string.dasherize
-    end
+    StimulusHelpers.dasherized_cache[string] ||= string.dasherize
   end

Test plan

• All 7 existing tests still pass

🤖 Generated with Claude Code