Consider using BigInt for the bencode integer token

[josecelano]

Relates to: #12 (comment)

In the new tokenizer, I decided:

• To use only byte vectors for integer and string tokens.
• Not to limit the size of the integer.

I wanted to keep this first tokenizer version as simple as possible, but we can discuss other @da2ce7's proposals like:

#[derive(Debug)]
pub enum Bencode<'a, R: BufRead + 'a> {
    Integer(BigInt),
    ByteString(Vec<u8>),
    List(BencodeListIterator<'a, R>),
    Dictionary(BencodeDictIterator<'a, R>),
}

• The implementation limits the Integer to 1024bit.