support concurrent requests

[cmdoret]

afaict, the shacl subprocess is currently a bottleneck, as it:

1. blocks execution
2. always use the same data/shapes filename

We could take advantage of fastapi's async capabilities by:

1. running shacl asynchronously (e.g. with asyncio.create_subprocess_shell and/or a queue)
2. using temporary files with random filenames (e.g. via tempfile)

This would allow handling multiple requests concurrently. In practice, this means the API server could be operational and receive data for next requests while shacl is still running.