Initial implementation of generators

[SemyonSinchenko]

The main idea is to allow use implemented in rust functions in any environment (locally via some python glue, on Apache Spark cluster via mapInArrow, etc.)

• Each function is related to the corresponding part of the original R code;
• Each function is seed-based, so running it multiple time with the same seed will constantly return the same result;
• Join-datasets generation contains also keys_seed that should be the same for the whole sessions (all the batches);
• To allow parallel generation / out-of-core generation, etc each function return a batch of the requested size instead of the whole data like in the original implementation in R;
• Input of the rust functions is assumed to be valid (like described in docstrings) just because it is much easier to validate inputs in python glue;

Python API, PySpark API, Deltalake API, etc. will be done after finalizing a concept in rust-part.