Is your feature request related to a problem? Please describe.
Most of the time, you want to train your model with all the samples (entities/ ids) in the dataset until a specific timeframe.

This is a significant and basic feature when training a model.

Describe the solution you'd like
The current behavior looks like this when you need to load all the item ids in the program:

First you need to load somehow all the item ids, which for now in Feast there must be a workaround to do it.

After we loaded all the item ids of all the entity into item_ids we can call for get_historical_features

item_entity_df = pd.DataFrame.from_dict(
   {
       'item_id': item_ids,
       'event_timestamp': [datetime(2025, 1, 1)] * len(item_ids)
   }
)

# retrive datasets for training
item_df = store.get_historical_features(entity_df=item_entity_df, features=item_service)

Note that the item_id entity must be declared; otherwise, an error will occur. In a real-world scenario, I would like to retrieve all item IDs up to a specific timestamp, for example:

item_entity_df = pd.DataFrame.from_dict(
   {
       'event_timestamp': datetime(2025, 1, 1)
   }
)

# retrive datasets for training
item_df = store.get_historical_features(entity_df=item_entity_df, features=item_service)

NOTE that we define only one timestamp and not a list, and we don't define item_id in the entity_df.

A much more intuitive way should be supported as well when we the code above can be looks like this:

# retrive datasets for training
item_df = store.get_historical_features(until_date=datetime(2025, 1, 1), features=item_service)

Or by default until_date = datetime.now()

# retrive datasets for training
item_df = store.get_historical_features(features=item_service)

In this way we have a super important method call, in very

Additional context
Thank you