Fix a flaky e2e test (#3183)

Fixes the `local_node_replace_order` e2e test. Currently, the whole
protocol stack is getting started in the beginning. After creating a
first order and before sending a replacement one, a new block is minted,
which starts the auction preparation mechanism. Sometimes, the auction
gets prepared before the replacement order is placed, making the test
fail. This PR removes this redundant block mining request.

A test run of the change:
https://github.com/cowprotocol/services/actions/runs/12502462800?pr=3183
(failed due to the GH action execution timeout; all the tests passed).

crates/e2e/tests/e2e/replace_order.rs

@@ -256,8 +256,6 @@ async fn single_replace_order_test(web3: Web3) {
     let balance_before = token_a.balance_of(trader.address()).call().await.unwrap();
     let new_order_uid = services.create_order(&new_order).await.unwrap();
 
-    onchain.mint_block().await;
-
     // Check the previous order is cancelled
     let old_order = services.get_order(&order_id).await.unwrap();
     assert_eq!(old_order.metadata.status, OrderStatus::Cancelled);