Harmonize on proto-types for representing well-known protobuf message types

[mzabaluev]

Version(s) of tendermint-rs: v0.23.2, master

Description

There is code in ibc-rs that jumps through hoops for the sole reason that the protobuf message structures generated for IBC use proto-types, while tendermint-proto includes a homespun variant of structures representing google.protobuf.

This was done in #639 which predates the rearchitecting with domain types and derives serde impls directly on the protobuf message types. This is less preferable than putting the serde concern cleanly onto the domain types, since only the latter can enforce value validity.

Definition of "done"

tendermint-proto uses proto-types to represent Timestamp, Duration, and possibly other well-known message types.
The corresponding domain types in tendermint respond to all serde needs.

[tony-iqlusion]

We've had some requests downstream in cosmos-sdk-proto for better alternatives for representing well-known types:

• Use Google's Protobuf well-known types definitions instead of prost_types cosmos/cosmos-rust#185
• prost-wkt-build to support Any::{try_pack, try_unpack} in cosmos-sdk-proto cosmos/cosmos-rust#176

The latter suggests using prost-wkt which is a richer crate for representing well-known types than prost-types and it also includes serde support.

It also has helper methods for packing and unpacking messages to/from an Any type which would be very helpful.

[mzabaluev]

Using "rich" crates to support non-protobuf functionality within prost-targeted types may not be the best long term approach. I've put some thoughts in #654 (comment)