Break apart Contents Abstraction

We currently have GithubRepo.get_contents as an abstraction around GitHub's Contents API endpoint. However, this endpoint behaves differently depending on what you're asking it for, in particular:

• a directory
• a file
• a file whose size is between 1MB and 100MB
• a file whose size is above 100MB

The method tries to handle those all in one place and the flow suffers because of it.

Instead we should pick apart this abstraction into two, based on the caller's intent:

• a directory
• a file, with fallbacks for larger files

For instance, file content for files between 1MB and 100MB can be retrieved with the application/vnd.github.v3.raw custom media type (as the Accept header). Since we already want the metadata from the initial call (it's not provided with a raw call) this seems like a nice fallback. We can discuss whether we need to support retrieving file content which is >100MB.

Start with GithubRepo.get_contents and the linked GitHub Contents API documentation, then trace its callers to distinguish directory and file intent. Define separate behavior for directories and files, including the raw-media fallback for files between 1MB and 100MB; the issue also requires deciding whether files over 100MB must be supported.