Originally, repository file lookups required a path within the

repository and a revision, which was fine for a lot of SCMs but wasn't

sufficient for others. Later, we added base_commit_id, a somewhat

loosely-defined identifier that represented some commit that the helped

locate the files on some SCMs/hosting services.
While working to integrate with another new SCM, we've found that these

three identifiers really aren't enough, and we need other information

parsed in a diff, made available through the DiffX parser and stored in

DiffSet.extra_data and FileDiff.extra_data. However, there was no

way of getting access to this information, and bolting on several new

parameters to get_file()/get_file_exists() (and SCM and hosting

service backends) seemed like an unsustainable approach.
This change introduces a new FileLookupContext object, which is a

container for contextual information that might be useful for file

lookups. It effectively replaces the base_commit_id and request

on the Repository functions, and adds extra_data attributes for the

FileDiff, DiffCommit, and DiffSet (or the Parsed* equivalents

during diff parsing/validation).
This gives implementations access to any metadata parsed from the diff,

along with the user accessing the repository, the active HTTP request,

and the base commit ID. In the future, we'll be able to add new context

(or even functions for permanently caching computed lookup state from

these functions) without having to funnel new parameters everywhere.
base_commit_id and request are now soft-deprecated. The Repository

functions still accept them and will pass them on, and the SCMTool

implementations still receive them. We'll probably want to include

deprecation warnings in a future major version.

Unit tests pass.
This is pending extensive testing with the development of a new SCM

integration.