Motivation

This PR fixes two issues:

1. Users are forced to provide --sender argument when the sender can be determined from keystore or hw wallet. (A better UX about wallet management #6034)
2. Users are forced to enter private keys and keystore password several times when running multi-chain scripts.

The idea was to rewrite MultiWallet into something that can be passed through script execution

Solution

This PR consists of several parts:

1. Refactor of foundry-wallets. Initially the plan was to refactor it into 2 distinct parts: logic related to user input (clap, validation, etc) and logic related to actual signing and wallets, but it seems that besides the former logic foundry-wallets is basically 100 lines of thin wrappers around ethers-signers. Not sure what's the best way to deal with those, so right now I just extracted all logic for reading wallets and creating signers into reusable free functions in wallets/utils, so that it's easier to navigate and we no more duplicating stuff.
2. Add MultiWallet and PendingWallet. MultiWallet contains set of WalletSigner which is our enum wrapping ethers-signers objects and also a set of PendingWallet, those are wallets which require user action to be unlocked. Currently it's only keystores and interactively requested private keys. The idea here is that MultiWallet is an entity created once and used for the whole duration of the wallet interaction and on the first time signers are requested, all pending wallets are getting unlocked.
3. To make signers independent from chain_id I changed our logic to completely ignore chain_id on signers and set it to 1 by default. This can be cleaned up once we migrate to alloy signers which have chain_id field optional
4. Add MultiWallet to cheatcodes inspector behing Arc<Mutex>. Use it in cheatcodes to determine sender for "anonymous" broadcasts and add remembered private keys directly to MultiWallet storage. Mutex here shouldn't really cause any threads to wait for locks as Cheatcodes are only getting cloned and used in several threads when running tests, for which script_wallets is being set to None as we are not collecting anything from it for tests.

Not sure if there was a demand for it, but this brings wallet-management closer to the cheatcodes, and would let us add support for vm.sign through local wallets (for example, signing stuff in script via ledger or keystore).