Fix naive requests usage

Currently schemas and robots are downloaded and cached without validating that they are loadable.

Solution: Use cobra-io to download -> will raise errors if impossible. Uses more request library to cache + react to timeouts.