Fix quoted_printable_encode() exceeding the line length for bytes 0xF5 to 0xFF

[auraseb]

php_quot_print_encode() decides whether to insert a soft line break based on the byte being encoded. Bytes up to 0x7F and the UTF-8 lead and continuation bytes up to 0xF4 are covered, but bytes 0xF5 to 0xFF match no branch, so they never cause a break. A run of such bytes (or binary data that ends a line with one) produces lines longer than the 76 characters RFC 2045 allows:

var_dump(strlen(quoted_printable_encode(str_repeat("\xFF", 30)))); // one 90 character line

These bytes can not start a UTF-8 sequence, so this treats them like single bytes and breaks as soon as the encoded byte no longer fits on the line.

The decoded output is unchanged; only the placement of soft line breaks differs. Added a test covering 0xF5, 0xF8 and 0xFF.

Targets the lowest supported bugfix branch; the patch applies unchanged there.

Fixes #24441