Unify DAG schedule args and change default to None

[uranusjr]

The arguments 'schedule_interval' and 'timetable' are removed from both the DAG class and the @dag decorator.

The default value of the 'schedule' argument (on both entities) is changed to None (i.e. a DAG will not have a schedule by default).

The 'timetable' attribute still exists on DAG, and is now the only value that reflects the DAG's schedule. The 'schedule_interval' attribute is removed from DAG.

The 'schedule_interval' on DagModel used to store a string representation of DAG's attribute of the same name, is now replaced by timetable_summary, which should (mostly?) work the same as before. We can fix minor UI differences as we go. Some use cases that rely on that field are also changed to use other fields instead (dataset_expression, for example, can be used to check whether a DAG is dataset-triggered).

The API field 'schedule_interval' has also been removed since that field no longer exists in the database. This has some side effects. The field previously contains some type information for delta values, but now becomes a simple string. It is unclear if anyone really needs the information, but we can always bring it back if needed.

Close #24842

[Lee-W]

One small nitpick. Everything else looks good to me, but I'm not an expert in the area. It would be great to have a second set of eyes.

[Lee-W]

if we're to add migration rule for this PR to #41641, we probably could do the following 2?

• schedule_interval -> timetable_summary
• schedule=NOTSET, schedule=None -> schedule=timedelta(days=1),

[uranusjr]

schedule_interval -> timetable_summary

These two are sort of interchangable for display purposes, but not if you want to use the value for something else. For example:

with DAG(schedule=timedelta(days=2)) as d1:
    ...

# This can not be changed to timetable_summary.
# d1.timetable would work, but fail for other use cases.
with DAG(schedule=d1.schedule_interval) as d2:
    ...

If we’re going the linter (ruff-like) route, this would be a case for showing a linting error without an automated fix available.

schedule=NOTSET, schedule=None -> schedule=timedelta(days=1),

schedule=NOTSET should be changed to timedelta(days=1), but schedule=None should be kept as-is (its behaviour is not changed in 3.0)