start: Pipelines trail

[iesahin]

This modifies start/data-pipelines for dvc stage add and a new project.

Closes #2857

[shcheklein]

Let's discuss first the detail in the #2857. Also, in terms of the priorities - I would say, we need to wrap up the experiments trail first.

[iesahin]

I modified this one not to use example-dvc-experiments, but example-get-started again. It currently uses only stage add but doesn't tell anything about running the pipeline. Which command is more appropriate here, dvc exp run or dvc repro? @shcheklein @dberenbaum

[dberenbaum]

For getting started, I would probably focus on exp run since it's already being introduced for experiments and may be more useful for the typical use case here. However, I do think there are situations where people may prefer repro, so we probably want to mention it briefly here and/or in more detail in the user guide.

I don't have a strong opinion, so interested to hear your thoughts.

[iesahin]

It may be better to provide a pipelines document without going into (a) metrics and plots (b) multiple stages.

[iesahin]

As we split stage definition and running them, the flow of text will also change.

[iesahin]

I think having two "under the hood" sections is unnecessary.

[iesahin]

I have added some notes to the content for discussion.

https://dvc-org-iesahin-issue28-s3ksse.herokuapp.com/doc/start/data-pipelines

@shcheklein @jorgeorpinel @dberenbaum

[dberenbaum]

At a quick glance, I think the discussion points make sense, although I wouldn't keep all examples in the ML/DS context.

[iesahin]

Moved this to Notion