test(g2p): fill expected phonemes for all 12 language variants

Expected values derived from the ChainG2p dict files and the Multig2p-Multi
config.json phoneme sets for each language:

  eng/default hello -> hh ah l ow  (CMU ARPABET, ah not ax)
  eng/plus   hello -> hh ax l ow  (ARPABET Plus has ax)
  deu/default  ja -> y aa         (OpenUTAU dict: y aa)
  fra/default  oui -> ou ii       (OpenUTAU dict: ou ii)
  ita/default  si -> s i          (remove_tone_digits strips stress)
  kor/default  가 -> g a           (Latin phoneme set)
  por/default  sim -> s i~        (nasal vowel i~)
  rus/default  да -> d a           (Latin phoneme set)
  spa/default  si -> s i
  fil/default  oo -> o o
  deu/marzipan ja -> j a          (IPA-like phoneme set)
  fra/millefeuille oui -> TBD     (different phoneme set, needs model)

Author: wolfgitpr

tests/g2p/tst_multig2p_languages.cpp

@@ -225,21 +225,34 @@ namespace {
 
     /// All 12 languages supported by Multig2p-Multi, each with the simplest
     /// common word to minimize ONNX variance. Language codes are ISO 639-3.
+    /// expectedPron values come from the dict training data (ChainG2p-*/...txt)
+    /// and config.json phoneme sets. For ita/default, preprocessor
+    /// remove_tone_digits strips stress marks (i1 → i).
     const std::vector<LangCase> &langCases() {
         static const std::vector<LangCase> cases = {
             // eng/default uses CMU ARPABET: ah (not ax) for schwa
             {"eng/default",       "hello", "eng (CMU)",          "hh ah l ow"},
             // eng/plus (ARPABET Plus) uses ax for the reduced schwa
             {"eng/plus",          "hello", "eng (ARPABET Plus)", "hh ax l ow"},
-            {"deu/default",       "ja",    "deu (OpenUTAU)",     ""},
-            {"fra/default",       "oui",   "fra (OpenUTAU)",     ""},
-            {"ita/default",       "si",    "ita (OpenUTAU)",     ""},
-            {"kor/default",       "가",    "kor (OpenUTAU)",     ""},
-            {"por/default",       "sim",   "por (OpenUTAU)",     ""},
-            {"rus/default",       "да",    "rus (OpenUTAU)",     ""},
-            {"spa/default",       "si",    "spa (OpenUTAU)",     ""},
-            {"fil/default",       "oo",    "fil (OpenUTAU)",     ""},
-            {"deu/marzipan",      "ja",    "deu (Marzipan)",     ""},
+            // deu dict: ja → y aa; voc has y, aa (OpenUTAU format)
+            {"deu/default",       "ja",    "deu (OpenUTAU)",     "y aa"},
+            // fra dict: oui → ou ii; voc has ou, ii
+            {"fra/default",       "oui",   "fra (OpenUTAU)",     "ou ii"},
+            // ita dict: si → s i1; remove_tone_digits → s i
+            {"ita/default",       "si",    "ita (OpenUTAU)",     "s i"},
+            // kor phoneme set uses Latin: syllable 가 → g a
+            {"kor/default",       "가",    "kor (OpenUTAU)",     "g a"},
+            // por dict: sim → s i~; voc has i~
+            {"por/default",       "sim",   "por (OpenUTAU)",     "s i~"},
+            // rus dict uses Latin phonemes: да → d a
+            {"rus/default",       "да",    "rus (OpenUTAU)",     "d a"},
+            // spa dict: si → s i
+            {"spa/default",       "si",    "spa (OpenUTAU)",     "s i"},
+            // fil dict has only uppercase OO → Q O Q O; lowercase oo → o o
+            {"fil/default",       "oo",    "fil (OpenUTAU)",     "o o"},
+            // deu/marzipan uses IPA-like voc: ja → j a
+            {"deu/marzipan",      "ja",    "deu (Marzipan)",     "j a"},
+            // fra/millefeuille uses different phoneme set; TBD from model
             {"fra/millefeuille",  "oui",   "fra (Millefeuille)", ""},
         };
         return cases;