Asynchronously executing custom graphs

I'm trying to run multiple custom graphs via `client.get(...)` simultaneously but I'm having issues getting this to work.

The native `Futures` interface does not seem to support custom graphs but only executing individual functions.
I've also tried various permutations of the examples provided on the [Asynchronous Operation](https://distributed.dask.org/en/latest/asynchronous.html) section of the documentation (both in the `asyncio`and `torando` flavor), but to no avail.

The following code for instance yields an exception or cancels the tasks depending on where you apply the `asynchronous=True` argument:
```python
import asyncio
import time

from dask.distributed import LocalCluster, Client

def mywait(seconds: int):
print(f"I'm waiting for {seconds} seconds")
time.sleep(seconds)
return seconds

def passthrough(arg):
return arg

def run_asyncio():
cluster = LocalCluster(dashboard_address=22222, processes=True)
# client = Client(cluster)

graph_1 = {"wait 10": (mywait, 10), "end": (passthrough, "wait 10")}
graph_2 = {"wait 20": (mywait, 20), "end": (passthrough, "wait 20")}

async def _run_multiple_graphs(graphs):
client = await Client(cluster, asynchronous=True)
result = await asyncio.wait([client.get(graph, "end") for graph in graphs])
return result

output = asyncio.get_event_loop()\
.run_until_complete(_run_multiple_graphs([graph_1, graph_2]))
print(output)

if __name__ == "__main__":
run_asyncio()
```
As I will be starting computation of the graphs simultaneously a viable workaround is to just cleverly merge `graph_1` and `graph_2` into a single definition but I'm wondering whether that is the only solution.