Fix max_encoded_len for Compact fields

[pgherveou]

Noticed that when using compact fields the max_encoded_len macro expand to the same as if we were not using the compact attribute.

#[derive(Encode, Decode, MaxEncodedLen)]
pub struct Weight {
    #[codec(compact)]
    ref_time: u64,
    #[codec(compact)]
    proof_size: u64,
}

Will expand to

    impl ::codec::MaxEncodedLen for Weight {
        fn max_encoded_len() -> ::core::primitive::usize {
            0_usize
                .saturating_add(<u64>::max_encoded_len())
                .saturating_add(<u64>::max_encoded_len())
        }
    }

This PR fix it to

    impl ::codec::MaxEncodedLen for Weight {
        fn max_encoded_len() -> ::core::primitive::usize {
            0_usize
                .saturating_add(::parity_scale_codec::Compact<u64>::max_encoded_len())
                .saturating_add(::parity_scale_codec::Compact<u64>::max_encoded_len())
        }
    }

[pgherveou]

also a few side thoughts:
Do we really need saturating_add for implementing this trait?
Could this be a candidate for const fn?

[bkchr]

also a few side thoughts: Do we really need saturating_add for implementing this trait?

I mean we probably never use more than usize, but there is also no harm in using saturating_add.

Could this be a candidate for const fn?

const fn trait functions are not supported right now.

[bkchr]

Some nitpicks, otherwise looks good 👍

Good work :)

[bkchr]

@pgherveou could you release v3.6.5?

[pgherveou]

isnt 3.6.5 the current version, you mean 3.6.6?