[NFC][libc] Clarifies underscores in n-char-sequence. (#102193)

The C standard specifies
  n-char-sequence:
    digit
    nondigit
    n-char-sequence digit
    n-char-sequence nondigit

nondigit is specified as one of:
  _ a b c d e f g h i j k l m
    n o p q r s t u v w x y z
    A B C D E F G H I J K L M
    N O P Q R S T U V W X Y Z

This means nondigit includes the underscore character. This patch
clarifies this status in the comments and the test.

Note C17 specifies n-char-sequence for NaN() as optional, and an empty
sequence is not a valid n-char-sequence. However the current comment has
the same effect as using the pedantic wording. So I left that part
unchanged.

Author: mordante

libc/src/__support/str_to_float.h

@@ -1159,13 +1159,11 @@ LIBC_INLINE StrToNumResult<T> strtofloatingpoint(const char *__restrict src) {
       index += 3;
       StorageType nan_mantissa = 0;
       // this handles the case of `NaN(n-character-sequence)`, where the
-      // n-character-sequence is made of 0 or more letters and numbers in any
-      // order.
+      // n-character-sequence is made of 0 or more letters, numbers, or
+      // underscore characters in any order.
       if (src[index] == '(') {
         size_t left_paren = index;
         ++index;
-        // Apparently it's common for underscores to also be accepted. No idea
-        // why, but it's causing fuzz failures.
         while (isalnum(src[index]) || src[index] == '_')
           ++index;
         if (src[index] == ')') {

libc/test/src/stdlib/strtof_test.cpp

@@ -200,7 +200,7 @@ TEST_F(LlvmLibcStrToFTest, NaNWithParenthesesValidSequenceInvalidNumberTests) {
   run_test("NaN(1a)", 7, 0x7fc00000);
   run_test("NaN(asdf)", 9, 0x7fc00000);
   run_test("NaN(1A1)", 8, 0x7fc00000);
-  run_test("NaN(why_does_this_work)", 23, 0x7fc00000);
+  run_test("NaN(underscores_are_ok)", 23, 0x7fc00000);
   run_test(
       "NaN(1234567890qwertyuiopasdfghjklzxcvbnmQWERTYUIOPASDFGHJKLZXCVBNM_)",
       68, 0x7fc00000);