Should $Xmeta output be None-grouped

Consider something like this:

struct S {
    #[deftly(value_for_sum = "self.normal")]
    normal: u32,
    #[deftly(value_for_sum = "self.shifted << 1")]
    shifted: u32,
}
let sum = derive_deftly! {
    $( ${fmeta(value_for_sum)} + ) 0
};

This would expand to:

self.normal + self.shifted << 1 + 0

which is

(self.normal + self.shifted) << (1 + 0)

macro_rules! uses none-delimited groups to address this. We probably should too.

I need to do some more experiments. In particular, I'm not sure if this should apply to ... as ty.