Dataset versioning

[eedorenko]

1. A pipeline has the following parameters added:

• caller_run_id - for tracking purposes to identify who called the pipeline (e.g. ADF pipeline run id)
• dataset_version - makes it possible to rerun the pipeline with a specific dataset version. By default "latest"
• data_file_path - if provided, a pipeline registers a new version of the dataset pointing to the data_file_path

2. It is trackable now what dataset version (pointing to specific data file) was used to train the model
3. It is trackable now what models were trained with a given dataset version