Dataframe: re-design UI based on new query model

[abey79]

The goal of this redesign is to keep up with the changes being introduced in the data APIs #7455.

Overview: the is organised as a fixed pipeline of optional stages, where each stage corresponds to some primitive of the data API.

Stages:

• View (required): the subset of the datastore on which the rest of the filtering/querying happens.
  • content: entity path filter
  • timeline: the timeline used by everything else
• Filter: time range (optional): only keep rows whose timestamps on the view timeline are within the provided range.
  • start: low bound, can be set to -inf
  • end: high bound, can be set to +inf
• Filter: events (optional): only keep rows which contains a logging event for the specified column (formerly known as PoV).
  • entity: the column entity (must be one of the view's entity)
  • component: the column component (must be on of the specified entity)
• Fill with latest-at (bool, optional): whether or not null cells are filled using latest-at queries.
• Select (list of columns): for each view column, whether or not it is included or not in the final dataframe.
  • Helper to select all or nothing, or a recommended subset would be nice.

Notes:

• All components for the view entities are included in the view. It is not possible to further restrict this (like you can with the API).
• The dataframe view doesn't allow combing multiple filter_events_for. There is 0 or 1 active at a time.
• The dataframe view doesn't expose anything about custom time values (e.g. filter_time, latest_at_times).