Accumulate loop dyn size

[albertz]

Support of accumulation of dynamic sizes (DimensionTag.dyn_size_ext) within a recurrent context (control_flow_context). E.g. when the dyn size (seq lengths) is of shape [B] inside the loop, and this dyn size changes in every iteration (i.e. control_flow_context is set to the rec loop), then the accumulated dyn size outside the loop would have shape [T,B].

Further, when some layer inside the loop has such a dynamic shape [B,T_per_frame], where T_per_frame is the dyn size as mentioned before, this means that we cannot use the current TensorArray logic, which expects that elements always have the same shape in every frame. This PR also extends that logic to support that.

This was a request by #635 and also makes sense for #589, #579 (and other related commits for #391).

[albertz]

Finally. Again more effort than expected. This still needs to be cleaned up. Maybe some aspects will be separated. Like the DimensionTag.declare_same_as changes.