Optimize tar::archive::EntriesFields::next_entry_raw()

[jiangliu]

tar::archive::EntriesFields::next_entry_raw() has a performance bottleneck caused by the code:

        // Make sure the checksum is ok
        let sum = header.as_bytes()[..148]
            .iter()
            .chain(&header.as_bytes()[156..])
            .fold(0, |a, b| a + (*b as u32))
            + 8 * 32;

[jiangliu]

The code at https://github.com/alexcrichton/tar-rs/blob/master/src/archive.rs#L260 may be optimized as:

        // Make sure the checksum is ok
        let data = header.as_bytes();
        let mut sum = 8 * 32;
        for idx in 0..148 {
            sum += data[idx] as u32;
        }
        for idx in 156..512 {
            sum += data[idx] as u32;
        }