[feature] add general mcts agent

[TataKKKL]

rename mcts_agent to rmcts_agent to keep the original implementation
add new mcts with new agent config

[TataKKKL]

node expansion: multiple run to get more candidates

[TataKKKL]

(1) node expansion
talking about expansion, what I implemented right now is that children nodes are just generated all at once, based on trajectory from root to current node
however, I just realize, sometimes, just based on the trajectory information, we can only generate limited number of children nodes, but through trial and error (which means not only the trajectory but also some other trial and error information), we can some better nodes
then need to change the tree structure as well
let me think about how to rephrase it better, or think if it is really an issue now
for easy test cases, this really doesn't matter, but I feel this matters to difficult test cases

#15

[TataKKKL]

(2) Extend current LATS to make it customizable in Tree Policy and Rollout Policy
Modularize current LATS for more customized search setup options

extend it from “vanilla” UCT to more bandit-style scores such as Thompson Sampling, Policy-guided UCT
enable different rollout such as ε-greedy rollout instead of random rollout

#21

[TataKKKL]

(3) add MCTSrAgent
Adapting https://arxiv.org/pdf/2406.07394
#16