Incorrect processing of header's value

[DmitriyMaksimov]

If we feed multipart parser 1 byte in a time (call multipart_parser_execute() with len == 1) following code:

  case s_header_value:
    multipart_log("s_header_value");
    if (c == CR) {
      EMIT_DATA_CB(header_value, buf + mark, i - mark);
      p->state = s_header_value_almost_done;
    }
    if (is_last)
        EMIT_DATA_CB(header_value, buf + mark, (i - mark) + 1);
    break;

call on_header_value callback twice: first time with len == 0 (which is feature - indicates end of value) and second time with len == 1 and buffer[0] == CR which is bug.
Looks like we missing else statement:

    multipart_log("s_header_value");
    if (c == CR) {
      EMIT_DATA_CB(header_value, buf + mark, i - mark);
      p->state = s_header_value_almost_done;
    } else if (is_last)
        EMIT_DATA_CB(header_value, buf + mark, (i - mark) + 1);

or break:

    multipart_log("s_header_value");
    if (c == CR) {
      EMIT_DATA_CB(header_value, buf + mark, i - mark);
      p->state = s_header_value_almost_done;
      break;
    }
    if (is_last)
        EMIT_DATA_CB(header_value, buf + mark, (i - mark) + 1);

[vincentbernat]

This bug seems to be fixed in #19.