Async runtimes / parallelization in atomic-lib

[joepio]

atomic-server does a lot of async stuff, as an Actix project.

However, atomic-lib does not. @AlexMikhalev recently added paralellization to Db::all_resources, which helped speed it up. There are probably other places, too, where it might help to spin up some threads.

I'm not sure where we can expect to gain a lot by paralellization, but I expect that Db::query and Commit::apply_opts as costly functions that do a lot of loopy stuff could speed up quite a bit.

[AlexMikhalev]

Add benchmarks to measure the performance of specific queries and I will have a go at speeding it up - let's chunk this task into multiple subtasks.

[AlexMikhalev]

We have to rethink async and parallelisation in connection with the discussion https://github.com/atomicdata-dev/atomic-data-rust/issues/433 - Sled:IVec doesn't have Sync implemented so can't scale reads via Arc, we can use https://crates.io/crates/dashmap as cache, serialize it to disk using serde and then at some point migrate from sled if or when we need to. I also saw the news that sled is sunsetting, but I haven't seen confirmation.

For parallel we need to make a choice: are we going to add tokio as a dependency for atomic-data-rust or we need to make calls to parallelized functions - like all_resources sync from atomic-data-server. You can't mix both - tokio async and any other threadpool executors. It's possible to use tokio for CPU-bound tasks, see here and here but it requires dedicated effort.