Document how to use `open_root_actor()`

[goodboy]

With #185 about to land most users won't have to know about open_root_actor() however there is at least one use I can think of that might require it.

If you want to start a lone root actor that has no children but that is a long running service which can spawn workers (this also pertains to our standalone example which needs to be updated to remove tractor.run() as well) then you will need open_root_actor() to do this.

Likely code in that case is something close to:

import trio
import tractor

log = tractor.get_logger(__name__)

@tractor.stream
async def spawn_and_do_work(ctx: tractor.Context, numbers: List[int]):

    async with tractor.open_nursery() as tn:
    
        for n in numbers:
        
            # spawn an actor per entry in input sequence
            # (normally you'd probably want to retrieve results async too)
            result = await tn.run_in_actor(is_prime, num=number)
            
            # stream each result to caller
            await ctx.send_yield(result)


async def main():
    async with tractor.open_root_actor(expose_modules=[__name__]) as actor:
        log.info("yo i'm the root actor")
        await trio.sleep_forever()
    
trio.run(main)

Then, in another actor tree program you'd find and call this actor to do spawn_and_do_work() as needed but,

1. each time a new request comes in open_root_actor() isn't called implicitly over and over inside the root nursery
2. the actor machinery is up making it possible to find this actor in the first place using the registry system