[Here](https://github.com/JuliaReinforcementLearning/ReinforcementLearningCore.jl/blob/a9fcedc25d0c1e98a2ed671bfeed822b850bb14a/src/utils/base.jl#L58), it is mentioned that ``` reshape(x, (size(x) |> reverse |> Base.tail |> Base.tail |> reverse)..., :) ``` is much faster than ``` reshape(x, size(x)[1:end-2]..., :) ``` A brief explanation as to why that is so would be really appreciated! Thanks.