Refactor "streams" to "real-time tables"

[begelundmuller]

Summary

Naming things is hard. One major term that's been aching for a while are our use of the word "stream". In Beneath, a stream is both a log with infinite retention, an indexed table, and a data warehouse table. The term "stream" is particularly bad for batch datasets, like uploaded dataframes ("batch stream", "finalized stream", ...).

I've considered and tested several alternative terms: tables, collections, topics, sets, datasets, frames, dataframes, slices, objects. We could also invent a new term altogether. But I think none of them beats the simplicity of "table", which everyone understands. By saying "real-time tables" or "subscribe to table" or "replay table", I think we can reasonably convey the log streaming feature.

So this refactor is about changing our use of "stream" across the code base with "(real-time) table".

Risks and challenges

Our APIs (gql and grpc) use "stream" extensively, and changing it will break old client versions. It's a good exercise to see if we're geared to maintain backwards compatibility.

Involved components

Virtually the entire codebase. This is a superficial change, but will touch every part of the codebase.

[begelundmuller]

Did the backend refactor: fb95180

[begelundmuller]

Did the frontend refactor: f089d0a

[begelundmuller]

Did the remaining clients, examples and docs refactors and pushed to production yesterday. Done!