[FLINK]Support function split_index

[KevinyhZou]

No description provided.

[shuai-xu]

Can we use unix_timestamp and split_part instead of adding two functions?

[KevinyhZou]

Can we use unix_timestamp and split_part instead of adding two functions?

unix_timestamp return type is bigint, can not be mapped to the flink function Proctime, it return type is timestamp;
split_part is a function of presto, and we have to register all the scalar functions of presto if we want to use it, and it will has some conflict with spark scalar functions, as I mentioned in pr: bigo-sg/velox4j#4