Why do strings have to be valid UTF-8?

Consider this example:

use std::io::Read;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut file = std::fs::File::open("number")?;
    let mut buf = [0_u8; 128];
    let bytes_read = file.read(&mut buf)?;

    let contents = &buf[..bytes_read];
    let contents_str = std::str::from_utf8(contents)?;
    let number = contents_str.parse::<i128>()?;

    println!("{}", number);
    Ok(())
}

Why is it necessary to convert the slice of bytes to an &str? When I run std::str::from_utf8, it will validate that contents is valid UTF-8. But to parse this string into an integer, I only care that each byte in the slice is in the ASCII range for digits as it will fail otherwise. It seems like the std::str::from_utf8 adds unnecessary overhead. Is there a way I can avoid having to validate UTF-8 for a string in a situation like this?

Edit: I probably should have mentioned that the file is a cache file I write to. That means it doesn’t need to be human-readable. I decided to represent the number in little endian. It should probably be more efficient than encoding to / decoding from UTF-8. Here is my updated code to parse the file:

use std::io::Read;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    const NUM_BYTES: usize = 2;

    let mut file = std::fs::File::open("number")?;
    let mut buf = [0_u8; NUM_BYTES];

    let bytes_read = file.read(&mut buf)?;
    if bytes_read >= NUM_BYTES {
        let number = u16::from_le_bytes(buf);
        println!("{}", number);
    }

    Ok(())
}

If you want to write to the file, you would do something like number.to_le_bytes(), so it’s the other way around.