Support context propagation without breaking changes

[goccy]

This PR is related to #925 .

First of all, thank you for the review the other day.
As suggested, I tried to avoid changing the public interface and instead added new types to support it. Although there are many changes, as you can see from the test code, I added the functionality without altering the existing behavior 🎉

Additionally, during the changes, it was unnecessary to pass context.Context to AddQualifierContext / QualifyContext / QualifyIfPresentContext, so if there are any issues, it's possible to revert these to use the original types. Please let me know your thoughts.

Pull Requests Guidelines

See CONTRIBUTING.md for more details about when to create
a GitHub Pull Request and when other kinds of contributions or
consultation might be more desirable.

When creating a new pull request, please fork the repo and work within a
development branch.

Commit Messages

• Most changes should be accompanied by tests.
• Commit messages should explain why the changes were made.

Summary of change in 50 characters or less

Background on why the change is being made with additional detail on
consequences of the changes elsewhere in the code or to the general
functionality of the library. Multiple paragraphs may be used, but
please keep lines to 72 characters or less.

Reviews

• Perform a self-review.
• Make sure the Travis CI build passes.
• Assign a reviewer once both the above have been completed.

Merging

• If a CEL maintaner approves the change, it may be merged by the author if
  they have write access. Otherwise, the change will be merged by a maintainer.
• Multiple commits should be squashed before merging.
• Please append the line closes #<issue-num>: description in the merge message,
  if applicable.

[TristonianJones]

@goccy thanks for taking another look at this. I'm sorry I haven't been able to review the change yet. It's a large change and I haven't had a sufficiently large block of time to review it in the manner it deserves to be. I will try to get to the review in more detail next week.

[TristonianJones]

@goccy I think I'm going to revive #368 as it will have fewer impacts and potential for breakages while still giving you most of what you want. The difference is that the implementation won't block on concurrent calls unless it needs to, instead iterating through calls in batches and re-evaluating. It's a definite overhead, but the highest cost should be the I/O rather than the repeated evaluation.