Feature Request: Add Debug bound to Error types for Database trait

[prestwich]

Currently downstream errors containing Db::Error have to add manual bounds for Debug if they want any printing strategy. However, Debug is derived or impld for all current errors used in revm. This would be a breaking change, however it is unlikely that any downstream implementors have non-Debug error types (I checked the foundry fork DB and they're already Debug)

This affects EVMError<Db::Error> because it derives Debug and therefore has no Debug impl unless the inner Db::Error has a Debug bound

affected traits:

• Database
• DatabaseRef
• State
• StateRef
• BlockHash
• BlockHashRef

What I want to do:

// broken code
#[derive(Debug, thiserror::Error)]
pub enum Error<Db> 
where
    Db: Database
{
    // results in an error 
    // `EvmError<Db::Error>` cannot be formatted using `{:?}` because it doesn't implement `Debug`
    #[error("{0:?}"]
    Evm(EVMError<Db::Error>),
    ...
}

What I have to do instead:

#[derive(Debug, thiserror::Error)]
pub enum Error<Db> 
where
    Db: Database
    // this bound holds for all current Database implementors, but must be manually written by consumers
    Db::Error: Debug,
{
    #[error("{0:?}"]
    Evm(EVMError<Db::Error>),
    ...
}

the same is true of std::error::Error, however:

• this approach does not dovetail with the crates extern crate alloc as std strategy for no_std support
• it would require an intermediary trait ErrBound to handle changing the bound for std/no_std
• there would need to be a cfg(feature = "std") implementation of std::error::Error for DatabaseComponentError

so while adding a std::error::Error it seems like supporting a bound of std::error::Error would be much more invasive, while adding Debug is non-invasive

[rakita]

This seems reasonable

We are currently in the process of bigger refactor on Generics here: #1672
So I will plan to add those at a similar time.