Improve let-chain indentation

Current formatted code:

fn too_long_function_name_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxcxxxxxxxxxxxxxxxxxxxx()
-> Option<i32> {
    None
}

fn main() {
    if let Some(x) =
        too_long_function_name_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxcxxxxxxxxxxxxxxxxxxxx()
        && x != 0
    {
        println!("x")
    }
}

If you don't pay attention to the = at the end of the previous line,
you may think that this call returns a bool

Expect improved formatted code:

fn too_long_function_name_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxcxxxxxxxxxxxxxxxxxxxx()
-> Option<i32> {
    None
}

fn main() {
    if let Some(x)
        = too_long_function_name_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxcxxxxxxxxxxxxxxxxxxxx()
        && x != 0
    {
        println!("x")
    }
}

or

fn too_long_function_name_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxcxxxxxxxxxxxxxxxxxxxx()
-> Option<i32> {
    None
}

fn main() {
    if let Some(x) =
            too_long_function_name_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxcxxxxxxxxxxxxxxxxxxxx()
        && x != 0
    {
        println!("x")
    }
}

Reproduce the issue with the two Rust examples in the report, then locate the rustfmt code path responsible for formatting let-chain conditions. Decide which of the two proposed layouts is intended and add a regression test for it. Done means the formatter produces the chosen improved indentation without regressing related let-chain formatting.