Minor: improve performance of ScalarValue::Binary* debug

[alamb]

Which issue does this PR close?

Rationale for this change

This is similar to #12322

(Another) Follow on to #12192 from @lewiszlw

What changes are included in this PR?

1. Add tests for binary debug
2. Fix a bug that the tests found
3. improve performance with fewer allocations

Are these changes tested?

Yes

Are there any user-facing changes?

[alamb]

this is a bug (likely copy/paste error) found with the tests

[alamb]

the difference here is that join allocates a new String

[Dandandan]

... and format! as well :)

[lewiszlw]

👍

[Dandandan]

I think this is slightly more idiomatic written as follows:

fn fmt_binary(data: &[u8], f: &mut fmt::Formatter) -> fmt::Result {
    let mut data_iter = data.iter();
    if let Some(b) = data_iter.next() {
        write!(f, "{}", b)?;
    }
    for b in data_iter {
        write!(f, ",{b}")?;
    }
    Ok(())
}

[Dandandan]

Or even better

fn fmt_binary(data: &[u8], f: &mut fmt::Formatter) -> fmt::Result {
    if let Some(b) = data.first() {
        write!(f, "{}", b)?;
    }
    for b in data.iter().skip(1) {
        write!(f, ",{b}")?;
    }
    Ok(())
}

[alamb]

I came up with something slightly different that I still think is more idiomatic in 99a57e7.

/// writes a byte array to formatter. `[1, 2, 3]` ==> `"1,2,3"`
fn fmt_binary(data: &[u8], f: &mut fmt::Formatter) -> fmt::Result {
    let mut iter = data.iter();
    if let Some(b) = iter.next() {
        write!(f, "{b}")?;
    }
    for b in iter {
        write!(f, ",{b}")?;
    }
    Ok(())
}

[Dandandan]

Added one small suggestion