Create a JSON IO Manager

[zaneselvans]

Overview

• Currently we generate a number of datapackage.json files as well as some xbrl_taxonomy.json outputs.
• Writing and reading these files is all managed manually, rather than running it through Dagster, so we're constantly sneaking around on the filesystem and indepednently need to make sure that the files are ready by depending on some other asset that isn't actually what we want.
• Instead we could create a pretty simple JSON IO Manager that makes Dagster aware of how to write/read these assets and lets us centralize and standardize how we handle them.
• E.g. some of the "enrichments" that happen at run-time in the new pudl_datapackage asset could just be a standard part of any datapackage.json descriptor that annotates a directory full of parquet files.