touch: Italian, Spanish and Romanian keyboard layouts keep the digits (#747)

The physical-keyboard remap for these layouts put accented letters on the
digit keys: with Italian on, Sym+W/E/R on a T-Deck Plus typed a, e and i with
graves instead of 1, 2 and 3, and the shifted-symbol fallback turned ! @ #
into their capitals. The T-Deck has no number row, so the Sym layer's digits
were the only ones there were. Those letters all come from the accent box
after their base letter (on by default), so the digit rows are now
pass-through. Latvian and the non-Latin layouts keep theirs: their letters
have no other way in.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

src/ui-touch/KeyboardLayouts.cpp

@@ -874,8 +874,10 @@ static const char* hw_nl_upper[26] = {
 static const char* hw_nl_digits[10]       = { nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr };
 static const char* hw_nl_digits_shift[10] = { nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr };
 
-/* Spanish stays close to US QWERTY; use the number row to expose n-tilde on
- * compact hardware where the semicolon key is absent. */
+/* Spanish stays close to US QWERTY. The digits stay digits (#747): n-tilde comes
+ * from the accent box after an 'n', like every other accented Latin letter, and on
+ * the T-Deck the only digits there are the Sym layer's, so a remap took them away
+ * (and '!' with them, through the shifted-symbol fallback below). */
 static const char* hw_es_lower[26] = {
     "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
     "k", "l", "m", "n", "o", "p", "q", "r", "s", "t",
@@ -886,10 +888,12 @@ static const char* hw_es_upper[26] = {
     "K", "L", "M", "N", "O", "P", "Q", "R", "S", "T",
     "U", "V", "W", "X", "Y", "Z"
 };
-static const char* hw_es_digits[10]       = { nullptr, "ñ", nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr };
-static const char* hw_es_digits_shift[10] = { nullptr, "Ñ", nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr };
+static const char* hw_es_digits[10]       = { nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr };
+static const char* hw_es_digits_shift[10] = { nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr };
 
-/* Italian uses the number row to surface the common accented vowels directly. */
+/* Italian keeps the digits (#747: Sym+W/E/R typed à è ì instead of 1 2 3, and
+ * '!' '@' '#' turned into À È Ì through the shifted-symbol fallback). The accented
+ * vowels come from the accent box after the vowel, as for Spanish and Romanian. */
 static const char* hw_it_lower[26] = {
     "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
     "k", "l", "m", "n", "o", "p", "q", "r", "s", "t",
@@ -900,11 +904,11 @@ static const char* hw_it_upper[26] = {
     "K", "L", "M", "N", "O", "P", "Q", "R", "S", "T",
     "U", "V", "W", "X", "Y", "Z"
 };
-static const char* hw_it_digits[10]       = { nullptr, "à", "è", "ì", nullptr, nullptr, nullptr, nullptr, nullptr, nullptr };
-static const char* hw_it_digits_shift[10] = { nullptr, "À", "È", "Ì", nullptr, nullptr, nullptr, nullptr, nullptr, nullptr };
+static const char* hw_it_digits[10]       = { nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr };
+static const char* hw_it_digits_shift[10] = { nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr };
 
-/* Romanian keeps the standard alpha matrix; the number row surfaces the five
- * special letters on compact hardware (same trick as Italian's accented vowels). */
+/* Romanian keeps the standard alpha matrix and the digits (#747, as Italian): the
+ * five special letters come from the accent box after a, i, s or t. */
 static const char* hw_ro_lower[26] = {
     "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
     "k", "l", "m", "n", "o", "p", "q", "r", "s", "t",
@@ -915,8 +919,8 @@ static const char* hw_ro_upper[26] = {
     "K", "L", "M", "N", "O", "P", "Q", "R", "S", "T",
     "U", "V", "W", "X", "Y", "Z"
 };
-static const char* hw_ro_digits[10]       = { nullptr, "ă", "â", "î", "ş", "ţ", nullptr, nullptr, nullptr, nullptr };
-static const char* hw_ro_digits_shift[10] = { nullptr, "Ă", "Â", "Î", "Ş", "Ţ", nullptr, nullptr, nullptr, nullptr };
+static const char* hw_ro_digits[10]       = { nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr };
+static const char* hw_ro_digits_shift[10] = { nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr,nullptr };
 
 /* Latvian keeps the standard a..z alpha matrix (identical to English, matching
  * the on-screen deck above); the number row surfaces the eight macron/cedilla