fix: Don't ignore post-messages that we have seen referenced by pre-message before

[Hocuri]

There is a bug in pre-messages:

• A pre-message adds an entry to the msgs table with the rfc724_mid of the post-message
• If the pre-message and post-message are fetched in separate cycles:
  • prefetch_should_download() returns false, because rfc724_mid_exists() returns something
  • so, the message is not added to download_later, and never automatically downloaded.

This PR fixes the test at 73a2c94, if one adds these lines:

diff --git a/deltachat-rpc-client/tests/test_something.py b/deltachat-rpc-client/tests/test_something.py
index 9f9a40e32..2a9ce407a 100644
--- a/deltachat-rpc-client/tests/test_something.py
+++ b/deltachat-rpc-client/tests/test_something.py
@@ -979,5 +979,7 @@ def test_large_message(acfactory) -> None:
     )
 
     msg = bob.wait_for_incoming_msg()
+    msgs_changed_event = bob.wait_for_msgs_changed_event()
+    assert msg.id == msgs_changed_event.msg_id
     snapshot = msg.get_snapshot()
     assert snapshot.text == "Hello World, this message is bigger than 5 bytes"

and the test reproduces the bug if one adds:

diff --git a/src/smtp.rs b/src/smtp.rs
index 0a4ef049a..e936662e5 100644
--- a/src/smtp.rs
+++ b/src/smtp.rs
@@ -511,6 +511,8 @@ pub(crate) async fn send_smtp_messages(context: &Context, connection: &mut Smtp)
 
     info!(context, "Selected rows from SMTP queue: {rowids:?}.");
     for rowid in rowids {
+        info!(context, "Sleeping before sending.");
+        tokio::time::sleep(std::time::Duration::from_secs(2)).await;
         send_msg_to_smtp(context, connection, rowid)
             .await
             .context("Failed to send message")?;

[iequidoo]

Just in case, this function can be implemented as rfc724_mid_exists_ex(context, rfc724_mid, &format!("download_state={}", DownloadState::Available)).is_some_and(|x| !x.1). Not sure if this is better.

Also could be renamed to rfc724_mid_download_tried.