Allow providers to define dynamic (inferred) timeouts

Introduce the ability for resources to define their timeouts dynamically, e.g. higher timeouts for EC2 instances of bigger instance sizes.

This way we could reduce the number of users which have to define timeouts and make their configs overall cleaner.

The issue names no files, tests, or entry points. Start by tracing how terraform-plugin-sdk resources currently define and resolve timeouts, then determine the provider-facing API and tests needed for size-dependent values; done means resources can infer timeouts from their configuration without requiring users to set them.

Written by the indexing model from the issue text.