On-demand tree loading

[philippotto]

For large tracings, downloading all trees can be a performance bottleneck.

Possible ideas to improve this:

• For a large amount of trees: Only download basic tree information and download the nodes of a tree only on demand (e..g, when activated)
• For very large trees: Split the tree in chunks (might be hard to find a plausible strategy here; maybe octtree-like splitting?)

The first scenario should be simpler to implement, which is why I'd argue to start with that.

Log Time

[fm3]

protoypical experiment: “main” tracing blob should hold index of existing trees when applying update actions, check which blobs need to be modified (possibly necessary to load main file) – update actions give tree id

test both extremes: save a blob for each tree vs hold everything in one file, (possibly filter content, and send only some trees to client)