Support predictions.create with model, version, or deployment parameters

[mattt]

This PR updates predictions.create to support overloads with model, version, or deployment parameters. With these changes, API consumers can more easily change between official models, model versions, and deployments.

import replicate

prediction = replicate.predictions.create(
    model="meta/meta-llama-3-8b-instruct",
    input={"prompt": "write a haiku about corgis"},
)

prediction = replicate.predictions.create(
    version="39ed52f2a78e934b3ba6e2a89f5b1c712de7dfea535525255b1aa35c5565e08b",
    input={"prompt": "a studio photo of a rainbow colored corgi"},
)

prediction = replicate.predictions.create(
    deployment="my-username/my-embeddings-model",
    input={"text": "hello world"},
)