Consider compatibility with borsh

[ARitz-Cracker]

Is adding #[derive(BorshSerialize, BorshDeserialize)] to all the structs provided by cosmwasm-std something you're willing to consider? Especially with a "2.0" coming out soon? I have a couple of contracts in development that are currently using bincode to serialize their state, and while I understand why you may want to serialize u64's and larger as strings when communicating with JS/TS systems, when it comes to storing variable-length data as cheaply as possible which is going to be served to the client through queries anyway, I would like a way to opt out of that.

As a sidenote, deriving Zeroable and Pod traits from bytemuck on fixed-length types (i.e. all the number types) could be cool as well.