Streamlining transformations for Split and Concat nodes.

[mdanilow]

New transformations in finn/transformation/streamline/reorder.py needed for yolov8 streamlining, some refactoring and generalization of existing transformations included.

• The MoveLinearPastEltwiseAdd() is now generalized by a refactored MoveIdenticalOpPastJoinOp and should be considered deprecated
• Some transformations (MoveScalarLinearPastInvariants, MakeMaxPoolNHWC, MakeScaleResizeNHWC) are fixed to check whether the node to be moved is a fork node, in which case the MoveOpPastFork would be called
• The MoveOpPastFork() transformation is more flexible because it uses deepcopies of an original node (the get_attrs_fxn parameter is not needed anymore)
• MoveScalarLinearPastSplit() and MoveTransposePastSplit() transformations added
• The MoveIdenticalOpPastJoinOp() transformation is more general and supports moving arbitrary number of identical operations past the join node (instead of just two). This base class checks only the op_types of producers to verify if they are identical. Its are_producers_identical() and move_node() methods should be overwritten by a child class for specific cases of producers and join op. Implemented child classes include: MoveTransposePastJoinAdd, MoveMulPastJoinAdd, MoveAddPastJoinAdd, MoveTransposePastJoinConcat, MoveAffinePastJoinConcat some needed only are_producers_identical() method to be overwritten.
• MoveAffinePastJoinConcat() transformation can be used for both scalar and channelwise affine operations.
• test_move_identical_op_past_join_op.py is removed and test_move_identical_op_past_join_add.py, test_move_identical_op_past_join_concat.py, test_move_identical_op_past_split.py are added instead

[auphelia]

If the new transformations replace this one, it can be deleted.