Stop using deprecated load_all_summary_data in everest_data_api.py:summary_values

[yngve-sk]

    def summary_values(self, batches=None, keys=None):

Currently it does this:

            summary = ensemble.load_all_summary_data()

Should use .load_response("summary", ...). Should do all the data operations with the polars dataframe before converting to pandas before return.

(It is possible to not convert to pandas at all if we rewrite everviz but that should probably be outside the scope of this issue)