regexp_nfa: Fix invalid fallthrough in character class detection

When the end character in a range matches a different standard range
(e.g., [0-z]), the range would be incorrectly detected as the class of
the end character (CLASS_az).

Instead of using a fallthrough, immediately FAIL when the end character
doesn't match the expected range.

src/nvim/regexp_nfa.c

@@ -634,6 +634,7 @@ static int nfa_recognize_char_class(char_u *start, char_u *end, int extra_newl)
           config |= CLASS_o7;
           break;
         }
+        return FAIL;
       case 'a':
         if (*(p + 2) == 'z') {
           config |= CLASS_az;
@@ -642,6 +643,7 @@ static int nfa_recognize_char_class(char_u *start, char_u *end, int extra_newl)
           config |= CLASS_af;
           break;
         }
+        return FAIL;
       case 'A':
         if (*(p + 2) == 'Z') {
           config |= CLASS_AZ;
@@ -650,7 +652,7 @@ static int nfa_recognize_char_class(char_u *start, char_u *end, int extra_newl)
           config |= CLASS_AF;
           break;
         }
-      /* FALLTHROUGH */
+        return FAIL;
       default:
         return FAIL;
       }

test/functional/legacy/036_regexp_character_classes_spec.lua

@@ -275,4 +275,16 @@ describe('character classes in regexp', function()
     diff(sixlines(string.sub(punct1, 1)..digits..punct2..upper..punct3..
       lower..punct4..ctrl2..iso_text))
   end)
+  it('does not convert character class ranges to an incorrect class', function()
+    source([[
+      1 s/\%#=0[0-z]//g
+      2 s/\%#=1[0-z]//g
+      3 s/\%#=2[0-z]//g
+      4 s/\%#=0[^0-z]//g
+      5 s/\%#=1[^0-z]//g
+      6 s/\%#=2[^0-z]//g
+    ]])
+    diff(string.rep(ctrl1..punct1..punct4..ctrl2..iso_text..'\n', 3)
+      ..string.rep(digits..punct2..upper..punct3..lower..'\n', 3))
+  end)
 end)