refactor: warn when we receive broken data from write buffer

We need to understand why this is the case and where broken data gets
stuck, because at the moment it seems that some Kafka hiccups can
poison the write buffer but it seems restarting IOx recovers the
situation.

server/src/write_buffer.rs

@@ -142,7 +142,7 @@ async fn stream_in_sequenced_entries<'a>(
             Ok(db_write) => db_write,
             // skip over invalid data in the write buffer so recovery can succeed
             Err(e) => {
-                debug!(
+                warn!(
                     %e,
                     %db_name,
                     sequencer_id,