Support for working with blueprint files

[jleibs]

Proposal for Simplification

Tasks 1-4 below can likely be done most expediently by building a parallel blueprint-loader mechanism instead of attempting to re-use DataSource. This loader would be almost strictly simpler than the current generalized RRD loader.

• Blueprints are small and bounded. It doesn't need to stream the blueprint in incrementally. The whole thing can be retrieved and then added to the store.
• For MVP, It doesn't need to support data-loaders.
• For MVP It doesn't need to support drag-and-drop.

The biggest complexity in the WIP draft is doing app-id-rewriting. For 1 this is a nice-to-have rather than required. We can always say your app-id must match to load the blueprint. However, if we want to support loading a blueprint in a differently named RRD, rather than rewrite app-id in the blueprint on load, we should be able to provide an API on the store-hub that allows cross-app-id association between blueprint and recordings.

Original Issue

• https://github.com/rerun-io/rerun/pull/5223/files is an initial implementation but required some ugly hacks.

This can probably be broken down into a few sub-steps:

• (1) File menu for save
• (2) File menu for load
• (3) Web-support for save/load
• (4) Web-support for url-param
• (5) Drag-and-drop support

These issues aren't directly blocking but are likely to be related:

• Don't activate blueprint until the entire file has been received #5297
• New options for setting blueprint: "force", "load if unset", "set default" #5298

[emilk]

For the url parameters, I think we should simply support a list of urls to load, which could be a .rrd/.rbl pair, or any amount of recordings and blueprints, or (in the future) other data too. This would make it more similar to just adding a lot of arguments to rerun-cli.

e.g. ?url=www.example.com/recording.rrd&url=www.example.com/blueprint.rbl

• EDIT: Location::query_map: support repeated key emilk/egui#4183