Consider getting rid of tokio

[emilk]

We only have a couple of crates using tokio. There are no benefits to it that I can see, but the costs are:

Requires a tokio runtime

We and users need to spawn a tokio runtime, e.g. using #[tokio::main] or tokio::runtime::Builder::new_multi_thread().

Forcing async onto users of the rerun crate is not very nice.

For us this is also annoying - if you forget to create a runtime you get a runtime crash if you call something requiring that runtime. Same for our users: the code may compile and run until they use a specific rerun feature (e.g. calling .serve())

async code is more complicated

And it requires special primitives, e.g. special "async" channels, when normal channels don't suffice.

Additional compilation times

It's not too bad, but is not nothing either:

tokio v1.28.1
├── bytes v1.4.0
├── libc v0.2.150
├── mio v0.8.11
│   ├── libc v0.2.150
│   └── log v0.4.17
│       └── cfg-if v1.0.0
├── num_cpus v1.15.0
│   └── libc v0.2.150
├── pin-project-lite v0.2.13
├── signal-hook-registry v1.4.1
│   └── libc v0.2.150
├── socket2 v0.4.9
│   └── libc v0.2.150
└── tokio-macros v2.1.0 (proc-macro)
    ├── proc-macro2 v1.0.78
    │   └── unicode-ident v1.0.8
    ├── quote v1.0.35
    │   └── proc-macro2 v1.0.78 (*)
    └── syn v2.0.48
        ├── proc-macro2 v1.0.78 (*)
        ├── quote v1.0.35 (*)
        └── unicode-ident v1.0.8
[build-dependencies]
└── autocfg v1.1.0