Async fetcher with limits

A price checker calls a dozen shop APIs. Each call spends nearly all its time waiting on the network, so instead of one after another, run them together on one asyncio event loop.

Implement the coroutine async def fetch_all(jobs: list) -> list.

• Each job is a zero-argument async function: await job() performs one fetch and returns its result.
• Start every job right away so they all wait on the network at the same time, then return their results in the same order as jobs, whatever order they finish in.
• fetch_all([]) returns [].
• Don't block the event loop: no time.sleep, no threads. The tests call asyncio.run(fetch_all(jobs)).

def make(value, delay):
    async def job():
        await asyncio.sleep(delay)
        return value
    return job

jobs = [make("a", 0.03), make("b", 0.01), make("c", 0.02)]
await fetch_all(jobs)     # ["a", "b", "c"] after about 0.03 s, not 0.06 s

The tests also count how many jobs are waiting at the same moment: with 50 jobs, all 50 should be in flight together.