Ignore currently running backup for SLA calculation

This was already the case (due to a bug) for a long time, but was
recently fixed in #37.
This commit reintroduces this behaviour.

The reasoning is that a backup which takes longer than 50% (grace
period) of the configured interval indicates a problem with the backup
server/network/etc...

changelog.d/20240612_121019_jb_running_backup_sla.rst

@@ -0,0 +1,3 @@
+.. A new scriv changelog fragment.
+
+- Ignore currently running backups for SLA calculation

src/backy/scheduler.py

@@ -91,8 +91,6 @@ def sla_overdue(self) -> int:
         """Amount of time the SLA is currently overdue."""
         if not self.backup.clean_history:
             return 0
-        if self.status.startswith("running"):
-            return 0
         age = backy.utils.now() - self.backup.clean_history[-1].timestamp
         max_age = min(x["interval"] for x in self.schedule.schedule.values())
         if age > max_age * 1.5:

src/backy/tests/test_daemon.py

@@ -250,6 +250,12 @@ def test_sla_over_time(daemon, clock, tmp_path, log):
     job.backup.scan()
     assert job.sla is False
 
+    # a running backup does not influence this.
+    job.update_status("running (slow)")
+    r = Revision.create(job.backup, {"daily"}, log)
+    r.write_info()
+    assert job.sla is False
+
 
 def test_incomplete_revs_dont_count_for_sla(daemon, clock, tmp_path, log):
     job = daemon.jobs["test01"]