refactor: make stricter database trait bounds for user ergonomics

[prestwich]

Two quick ergonomics changes for downstream users. Intended to simplify bounds writing

This is a breaking change, as downstream implementors of these traits may need to be updated

---

Error bounds for database errors

Adds an error bound for Database::Error and DatabaseRef::Error, and the async versions. This closes #1674. This prevents propagation of complex bounds for upstream wrapper structs, particularly error types that might wrap EVM output.

motivation:
simplify bounds writing for users by ensuring they don't need to manually write out error bounds. This is particulartly a problem when holding an instance of EVMError<Db>

Before:

use revm::{primitives::EVMError, Database};

pub struct Error<Db: Database<Error: core::error::Error>>(EVMError<Db::Error>);

impl<Db: Database<Error: core::error::Error>> From<EVMError<Db::Error>> for Error<Db>
{
    fn from(e: EVMError<Db::Error>) -> Self {
        Self(e)
    }
}

impl<Db: Database<Error: core::error::Error>> core::error::Error for Error<Db> {
    fn source(&self) -> Option<&(dyn Error + 'static)> {
        Some(&self.0)
    }
}

After:

use revm::{primitives::EVMError, Database};

pub struct Error<Db: Database>(EVMError<Db::Error>);

impl<Db> From<EVMError<Db::Error>> for Error<Db>
where
    Db: Database,
{
    fn from(e: EVMError<Db::Error>) -> Self {
        Self(e)
    }
}

impl<Db: Database> core::error::Error for Error<Db> {
    fn source(&self) -> Option<&(dyn Error + 'static)> {
        Some(&self.0)
    }
}

---

Add Database as a supertrait of DatabaseCommit

This does not close a current issue, but I'm happy to make one. This causes some bounds changes in the database_components examples

motivation: simplify bounds writing for users by allowing them to avoid specifying Database

alternate approach: a trait CommitableDatabase: Database + DatabaseCommit with a blanket impl<T> CommittableDatabase for T where T: Database + DatabaseCommit

Code before:

use revm::{Database, DatabaseCommit};
pub fn make_writable_evm<Db: Database + DatabaseCommit>(db: Db) -> Evm<'static, (), Db> { ... } 

Code after:

use revm::{DatabaseCommit};
pub fn make_writable_evm<Db: DatabaseCommit>(db: Db) -> Evm<'static, (), Db> { ... } 

---

Add WrapBlockHashRef to example

Add a wrapper struct like WrapDatabaseRef to BlockHashRef to allow relaxing the bounds on the database components in the database_components example

---

Drive-by:
The asyncdb is cfged out when running cargo c --all-features. And running cargo c -p revm-database-interface --features asyncdb produced compilation errors due to disabled tokio features. I have enabled the necessary features. When running with --workspace it appears some other package enabled the features, and covered up the dep specification issue

[prestwich]

hmmmm. for some reason it seems not all packages build when running cargo t on my local. so this may be more invasive than it appeared 🤔

punting to draft

[prestwich]

Ok it seems to be mostly an example that was broken, so I've updated that :)

I think this is ready for consideration. CI failure appears to be unrelated

[rakita]

Why 'static?

[prestwich]

in this case, it's required by the impl :Error for DatabaseComponentError<T, U> here. The bound could be removed here, but it would require all downstream users to propagate 2 bounds every time they want to use that impl Error for DatabaseComponentError (which is pretty often in practice)

Given that we don't have any examples of errors borrowing (i.e. all our current databases would meet this bound) it seems pretty reasonble to put it here too. This is just an example lib, and I don't feel strongly about its ergonomics tho

I would consider adding a + 'static bound to the Error assoc type in trait Database tho, for the same reason. We have no clear use case for errors that capture data instead of own it, and it causes a lot of downstream complexity

[rakita]

hm, i can't remember it but would this code

impl<T: BlockHashRef> BlockHash for T {

solve all of this

[prestwich]

it was unnecessary anyway so i removed it

[prestwich]

I would consider adding a 'static bound here, as described in https://github.com/bluealloy/revm/pull/1840/files#r1819000225

[prestwich]

I would consider adding a 'static bound here, as described in https://github.com/bluealloy/revm/pull/1840/files#r1819000225

[prestwich]

I would consider adding a 'static bound here, as described in https://github.com/bluealloy/revm/pull/1840/files#r1819000225

[prestwich]

I would consider adding a 'static bound here, as described in https://github.com/bluealloy/revm/pull/1840/files#r1819000225

[prestwich]

CI failure seems to be related to #1844

[prestwich]

closing as stale, may eventually resubmit