Sampler: add feature to generate samples without adding them to dataset

Feature request

We need a feature that generates random samples (single or multiple) of the design parameters but without adding them to the dataset.

Such a feature will be useful just to play around or to test the pipeline before running the expensive dataset generation procedure, for example in Grasshopper to look at ad hoc designs to see how they look like.

The feature could have a flag to specify output format. Minimal required options:

• 'dict' - dict in the format dict[objectname_key][nth_sample][ith_dimension]
• 'df' - for the flattened dataframe

Maybe in some pipelines the user would still want to write it to a file (without registering it in the dataset), but I guess this should be handled by the user separately.