Performance issues with current "contiguous" implementation

[kydos]

The implementation for the contiguous operation is sub-optimal for the case in which we have multiple slices.

This code:

zenoh/commons/zenoh-buffers/src/lib.rs

Line 105 in 7a47445

_ => Cow::Owned(slices.fold(Vec::new(), |mut acc, it| {

Should be rewritten to look something like:

_ => {
                    let mut l = 0;
                    for s in slices.by_ref() {
                        l += s.len();
                    }
                    let mut vec = Vec::with_capacity(l);
                    for slice in slices {
                        vec.extend_from_slice(slice);
                    }
                    Cow::Owned(vec)
                },

This will ensure a single allocation and will reduce to the minimum the total number of bytes actually copied.

[YuanYuYuan]

Has been completed via #1009 and #1022.