[FEA] Create a post block to be able to feed continuous list features to an MLP model

[rnyak]

Description

Currently, we are not able to feed continuous list features to an MLP model since the aggregation of this list features are not working as expected. We tested the following in the InputBlockV2 as continuous arg, but it does not work:

        continuous=mm.ContinuousFeatures.from_schema(schema_model.select_by_name(['length_of_stay_list']), 
                                                                           post=SequenceAggregator(SequenceAggregation.MEAN),
                                                                           aggregation='concat')

if we can make SequenceAggregator work for this case, that'd be solving the issue. It does not work now, since right now it has hard requirement that the input should be 3D. it should be able to work with 2D tensor as well.

This is an important feature for GTC'23 Merlin tutorial.

[rnyak]

fixed by #969