fix: Prevent long hiragana verb renyokei from absorbing onomatopoeia+する

Add penalty for 5+ character pure-hiragana verb renyokei not in
dictionary, mirroring the existing kPenaltyVeryLongHiraganaVerb logic.

Without this penalty, onomatopoeia followed by する conjugations (e.g.,
つるつるしている) could be parsed as a single spurious renyokei token
(つるつるし) instead of the correct split つるつる(ADV) + し(する).

The 15-byte threshold (5 hiragana chars * 3 bytes each) only applies to
VerbRenyokei to avoid penalizing legitimate long renyokei forms like
づけられる derived from dictionary verbs.

- Add regression test for 肌がつるつるしている in onomatopoeia.json

Author: libraz

src/analysis/scorer.cpp

@@ -467,6 +467,16 @@ float Scorer::wordCost(const core::LatticeEdge& edge) const {
     cost += sc::kPenaltyVeryLongHiraganaVerb;
   }
 
+  // Penalty for 5-char pure-hiragana verb renyokei not in dictionary
+  // E.g., "つるつるし" as godan-sa renyokei — should be つるつる(ADV) + し(する)
+  // Only renyokei: base forms like "づけられる" (from づける) are legitimate
+  if (!edge.fromDictionary() && edge.pos == core::PartOfSpeech::Verb &&
+      edge.extended_pos == core::ExtendedPOS::VerbRenyokei &&
+      grammar::isPureHiragana(edge.surface) &&
+      edge.surface.size() >= 15) {  // 5+ hiragana chars (5*3=15 bytes)
+    cost += sc::kPenaltyVeryLongHiraganaVerb;
+  }
+
   // Penalty for kanji+hiragana verb renyokei ending in いし pattern
   // E.g., "願いし" as renyokei of "願いす" is spurious
   // Should be 願い + し (願う renyokei + する renyokei)

tests/data/tokenization/onomatopoeia.json

@@ -959,6 +959,38 @@
         "cat_speech",
         "W-007"
       ]
+    },
+    {
+      "description": "onomatopoeia+する should split as つるつる+し, not つるつるし(verb)",
+      "expected": [
+        {
+          "pos": "Noun",
+          "surface": "肌"
+        },
+        {
+          "pos": "Particle",
+          "surface": "が"
+        },
+        {
+          "pos": "Adverb",
+          "surface": "つるつる"
+        },
+        {
+          "lemma": "する",
+          "pos": "Verb",
+          "surface": "し"
+        },
+        {
+          "pos": "Particle",
+          "surface": "て"
+        },
+        {
+          "pos": "Auxiliary",
+          "surface": "いる"
+        }
+      ],
+      "id": "gatsurutsurushiteiru",
+      "input": "肌がつるつるしている"
     }
   ],
   "description": "Onomatopoeia and mimetic words",