feat: Limit the size of aggregated WebXDC update to 100 KiB (#4825)

Before, update sending might be delayed due to rate limits and later merged into large
messages. This is undesirable for apps that want to send large files over WebXDC updates because the
message with aggregated update may be too large for actual sending and hit the provider limit or
require multiple attempts on a flaky SMTP connection.

So, don't aggregate updates if the size of an aggregated update will exceed the limit of 100
KiB. This is a soft limit, so it may be exceeded if a single update is larger and it limits only the
update JSON size, so the message with all envelopes still may be larger. Also the limit may be
exceeded when updates are sent together with the WebXDC instance when resending it as the instance
size isn't accounted to not complicate the code. At least this is not worse than the previous
behaviour when all updates were attached.

src/chat.rs

@@ -49,6 +49,7 @@ use crate::tools::{
     create_smeared_timestamps, get_abs_path, gm2local_offset, smeared_time, time, IsNoneOrEmpty,
     SystemTime,
 };
+use crate::webxdc::StatusUpdateSerial;
 
 /// An chat item, such as a message or a marker.
 #[derive(Debug, Copy, Clone, PartialEq, Eq)]
@@ -4272,9 +4273,39 @@ pub async fn resend_msgs(context: &Context, msg_ids: &[MsgId]) -> Result<()> {
         msg.timestamp_sort = create_smeared_timestamp(context);
         // note(treefit): only matters if it is the last message in chat (but probably to expensive to check, debounce also solves it)
         chatlist_events::emit_chatlist_item_changed(context, msg.chat_id);
-        if !create_send_msg_jobs(context, &mut msg).await?.is_empty() {
-            context.scheduler.interrupt_smtp().await;
+        if create_send_msg_jobs(context, &mut msg).await?.is_empty() {
+            continue;
         }
+        if msg.viewtype == Viewtype::Webxdc {
+            let conn_fn = |conn: &mut rusqlite::Connection| {
+                let range = conn.query_row(
+                    "SELECT IFNULL(min(id), 1), IFNULL(max(id), 0) \
+                     FROM msgs_status_updates WHERE msg_id=?",
+                    (msg.id,),
+                    |row| {
+                        let min_id: StatusUpdateSerial = row.get(0)?;
+                        let max_id: StatusUpdateSerial = row.get(1)?;
+                        Ok((min_id, max_id))
+                    },
+                )?;
+                if range.0 > range.1 {
+                    return Ok(());
+                };
+                // `first_serial` must be decreased, otherwise if `Context::flush_status_updates()`
+                // runs in parallel, it would miss the race and instead of resending just remove the
+                // updates thinking that they have been already sent.
+                conn.execute(
+                    "INSERT INTO smtp_status_updates (msg_id, first_serial, last_serial, descr) \
+                     VALUES(?, ?, ?, '') \
+                     ON CONFLICT(msg_id) \
+                     DO UPDATE SET first_serial=min(first_serial - 1, excluded.first_serial)",
+                    (msg.id, range.0, range.1),
+                )?;
+                Ok(())
+            };
+            context.sql.call_write(conn_fn).await?;
+        }
+        context.scheduler.interrupt_smtp().await;
     }
     Ok(())
 }