Avoid rate limiting by optimizing rpc calls

[0xzrf]

Description

The helper function create_accounts in tools-common/src/accounts_creator.rs:create_accounts() creates new accounts <= num_accounts argument provided to the function

The current functionality of the function is that it first gets rpc_client.get_latest_blockhash() on each loop and creates batches of transactions, and sending the txns to the rpc and then adding to the created_accounts vec

But design wise, it's not going to be able to avoid rate limit by the rpc based on the following observations:

1. Each loop calls the rpc_client.get_latest_blockhash() function, which is fine, but blockhash are valid for 150 slots(which is ~60 sec), more then enough to create all the accounts even if it's called once.
2. The create_transaction_batch function returns a Vec<Transactions>, each having a single system_program::create_account instruction. which are then sent to the send_transaction_batch function, which concurrently sends all the transactions(separately). imo, we can put up to 7 such instruction per transaction(adhering to 1232 bytes limit in a solana transaction). This will avoid rate limits presented by the rpc calls

[KirillLykov]

makes sense, I just haven't optimized this path.