Allow `connection_pool` to be set in RedisSettings

[brian-goo]

@samuelcolvin Thank you for a great library! It would be really grateful if this PR could be reviewed or a workaround for this issue provided. Many thanks.

---

We are using arq for our production app, and it appears that connection pool throws an error and arq worker crashes if the connection limit(max_connections) is reached. Not only does worker but the enqueue_job method also throws an error under the same condition.

This seems to be caused by the default ConnectionPool of redis-py.
https://github.com/redis/redis-py/issues/2517

Code to reproduce:

import asyncio
from arq import ArqRedis, create_pool
from arq.connections import RedisSettings

REDIS_SETTINGS = RedisSettings(
    max_connections=1
)

async def foo(ctx) -> None:
    print('ok')

async def main() -> None:
    redis: ArqRedis = await create_pool(REDIS_SETTINGS)
    for _ in range(2):
        await redis.enqueue_job('foo')

class Worker:
    functions = [foo]
    redis_settings = REDIS_SETTINGS

if __name__ == '__main__':
    asyncio.run(main())

By allowing connection_pool to be set in RedisSettings, it seems to work as expected.

import asyncio
from arq import ArqRedis, create_pool
from arq.connections import RedisSettings
from redis.asyncio import BlockingConnectionPool

REDIS_SETTINGS = RedisSettings(
    connection_pool=BlockingConnectionPool(max_connections=1)
)

async def foo(ctx) -> None:
    print("ok")

async def main() -> None:
    redis: ArqRedis = await create_pool(REDIS_SETTINGS)
    for _ in range(10):
        await redis.enqueue_job("foo")

class Worker:
    functions = [foo]
    redis_settings = REDIS_SETTINGS

if __name__ == "__main__":
    asyncio.run(main())

It appears that Worker can take redis_pool as an argument, but I guess we can handle this case more concisely this way for both Worker and Enqueuer, provided it does not break anything else.