Add feature transformation support to FeatureView about feast HOT 2 OPEN

I am starting to think having the decorator of feature view type is weird.

Instead we could just do something like:

@transform(
    sources=[credit_data_batch],
    entities=[user],
    mode="python",
    batch_schedule=timedelta(days=1),
    schema=[Field("user_id", String), Field("timestamp", Timestamp), Field("current_balance", Float64)],
)
def user_last_balance(transactions):
    return some_computation(transactions[["user_id", "timestamp", "current_balance"]])

And maybe in the decorator we have a way to map to when the computation is happening?

My view is that we need to provide clarity about when transformations happen (on demand, during a write, or in a stream--i.e., before a write) and I'm not sure the current way does this as obviously. Maybe it does.

from feast.

@franciscojavierarceo Just to make sure we're on the same page, do you think we should have a different BatchFeatureView class for this just like odfvs and stream feature views, right?

Also, I think we should probably implement it w/o scheduled materialization at first, because it's not really obvious to me which feast service could do the scheduling for this (flight server seems most logical or maybe we need to add another service altogether).

from feast.