Migrate remaining e2e test utilities to alloy primitives

[AryanGodara]

Description

Migrates internal ethcontract types to alloy primitives within the e2e crate, reducing reliance on legacy types.

Changes

• Replace ethcontract::primitives with alloy::primitives
• Simplify impersonate() function. Account return was unused; alloy handles impersonation via provider without wallet + .from(address); now returning unit (), since function's return value wasn't being used.
• Remove unnecessary into_legacy()/into_alloy() conversions

Note: ethcontract dependency remains for shared crate compatibility.

How to test

1. cargo check -p e2e --tests
2. Run e2e tests

[AryanGodara]

The side effect of anvil_impersonateAccount still happens
And since this return values wasn't actually being used where this function was called, I changed this to return unit type (also specified in PR description).

I could use some feedback on what should be the equivalent return type in alloyrs for ethcontract::Account, especially if the value isn't being used.

[MartinquaXD]

I think there are mostly 2 options:

1. impersonate() returns an alloy provider that doesn't have any signers (like here)
2. ForkedNodeApi has a impersonating_provider() that you are supposed to use for any number of different impersonated account.

I think I prefer 2 because it's pretty easy to communicate the correct way to use the API with doc comments.

#[derive(Debug, Clone)]
pub struct ForkedNodeApi {
    provider: AlloyProvider,
}

impl ForkedNodeApi {
    pub fn new(provider: AlloyProvider) -> Self {
        Self { provider: provider.without_wallet() }
    }

    /// Provider that can sign transactions on behalf of accounts
    /// previously impersonated with `[Self::impersonate()]`.
    pub fn impersonating_provider(&self) -> &AlloyProvider {
        &self.provider
    }

    /// Configures the node to sign any transaction for the given account.
    /// Such transactions should be sent from `[Self::impersonating_provider()]`.
    ///
    /// # Panics
    ///
    /// Panics if the connected RPC does not support impersonating account.
    pub async fn impersonate(&self, address: &Address);
    // all the other existing functions
}

Ultimately both versions are not optimal because you can set tx.from arbitrarily so nothing on the type level stops you from impersonating some account and then using any other impersonation incompatible provider to send transactions for that account.

[jmg-duarte]

You can use stop_impersonating_account — https://docs.rs/alloy-provider/latest/alloy_provider/ext/trait.AnvilApi.html#tymethod.anvil_stop_impersonating_account

The reason we need an empty wallet is that if the wallet has at least one signer, it will have a default signer, meaning all TXs will be signed using it. The alternative is to send the TX "manually" as in, outside the regular alloy pipeline that fills in gaps for the TX (like nonce, signature, etc)

Personally, I don't think the current state API is great, a better way of solving this would be having the impersonate taking an async future/closure that executes within a scope of impersonation:

Kinda like:

provider.impersonate(address, |empty_provider| async move {
	ERC20::instance(provider)::balanceOf(address)...
}).await

And inside, what really happens is:

self.transport
.execute("anvil_impersonateAccount", vec![json_address])
.await?;

let closure_result = closure.await;

self.transport
.execute("anvil_stopImpersonateAccount", vec![json_address])
.await?;

return closure_result

[jmg-duarte]

Or, better yet, remove this whole API and just use what anvil already provides: https://docs.rs/alloy-provider/latest/alloy_provider/ext/trait.AnvilApi.html#tymethod.anvil_send_impersonated_transaction_with_config

Maybe you can make it more ergonomic

[jmg-duarte]

Looking into this I opened #4004

Removed the forked node altogether, used alloy/anvil instead, no need to argue about the right return type :P

[MartinquaXD]

Probably makes sense to still address the impersonation topic in this PR.

[jmg-duarte]

LGTM asides the impersonation question, IMO we can either revert the change on the API and merge the rest of the PR; or fix it now

[jmg-duarte]

This function is unused, IMO we can remove this

[jmg-duarte]

Nerd nit: if you expand the macro you'll see that json! simply calls serde_json::to_value (at least for these two cases)

But it also makes the value you pass, a reference, meaning address becomes a double reference (&&Address)

Unlike other parts of the codebase, where address is immediately de-referenced turning it into a copy, here we can simply remove the macro (which usually hides more magic than in this case) use the expanded API and avoid a double reference

For the second call, we could do .as_str() or just the reference even.

I know this is not directly related to the migration, but a migration is the perfect opportunity to improve the codebase like this

[jmg-duarte]

Or, better yet, remove this whole API and just use what anvil already provides: https://docs.rs/alloy-provider/latest/alloy_provider/ext/trait.AnvilApi.html#tymethod.anvil_send_impersonated_transaction_with_config

Maybe you can make it more ergonomic