Add dynamic memory allocation option in astar_search method

[NorahXiong]

In A* search algorithm of freespace planner, memory of search space data is pre-allocated. If related parameters are different from default, it may result in high RAM usage. For example, if change "grid_resolution"(in file parking.launch.py) to 0.05, freespace planner will take up more than 16GiB RAM.
Actually, nodes in search space are sparsely accessed, so allocating memory for a node when it is accessed should be an option even though it will consume more time. I did some tests and results are as below. We can say most the pre-allocated memory is not used.

Case1:

Case2:

[NorahXiong]

Implemented in this PR.