deck server/client over a remote API

[Koeng101]

This PR adds a remote backend for the deck. This can be shared across many instances.

# server.py
import uvicorn
from pylabrobot.resources import Cor_96_wellplate_360ul_Fb, hamilton_96_tiprack_1000uL_filter
from pylabrobot.resources.hamilton import STARDeck
from pylabrobot.resources.remote.server import create_app

deck = STARDeck()
deck.assign_child_resource(hamilton_96_tiprack_1000uL_filter(name="tips"), rails=3)
deck.assign_child_resource(Cor_96_wellplate_360ul_Fb(name="plate"), rails=9)

app = create_app(deck)
uvicorn.run(app, host="0.0.0.0", port=8080)

# client.py
import asyncio
from pylabrobot.liquid_handling import LiquidHandler
from pylabrobot.liquid_handling.backends import STARBackend
from pylabrobot.resources.remote.client import RemoteDeck

deck = RemoteDeck.connect("http://localhost:8080")
lh = LiquidHandler(backend=STARBackend(), deck=deck)
await lh.setup()

tips = deck.get_resource("tips")
plate = deck.get_resource("plate")
await lh.pick_up_tips(tips["A1:H1"])
await lh.aspirate(plate["A1:H1"], vols=[100.0] * 8)
await lh.dispense(plate["A2:H2"], vols=[100.0] * 8)
await lh.drop_tips(tips["A1:H1"])

The key difference is that the deck can now be accessed through a standard API. This can be just in-memory (as it is right now), but it is now trivial to add a custom backing - whether that be SQL or document store or JSON files or anything else.

The concept is that you can have a server directly connected to a robot controlling it, but have the resource model be on a centralized server for state management.