perf(history_by_block_hash): reuse provider instead of creating two

[hai-rise]

Towards #12838.

Currently, history_by_block_hash creates a second provider instead of reusing a created one, which unnecessarily clones a bunch and creates an extra DB transaction. This is the biggest allocation per bytehound and a suspect that causes OOM when we spam submit 20-30M transactions.

[mattsse]

ah nice, yes definitely

[mattsse]

btw @hai-rise a similar problem exists here:

reth/crates/transaction-pool/src/validate/eth.rs

Lines 66 to 83 in fb64997

	pub fn validate_one(
	&self,
	origin: TransactionOrigin,
	transaction: Tx,
	) -> TransactionValidationOutcome<Tx> {
	self.inner.validate_one(origin, transaction)
	}
	/// Validates all given transactions.
	///
	/// Returns all outcomes for the given transactions in the same order.
	///
	/// See also [`Self::validate_one`]
	pub fn validate_all(
	&self,
	transactions: Vec<(TransactionOrigin, Tx)>,
	) -> Vec<TransactionValidationOutcome<Tx>> {
	transactions.into_iter().map(|(origin, tx)| self.validate_one(origin, tx)).collect()

this could use the same provider for all txs, could easily do this if we flip these

[mattsse]

let me open an issue for this