Spike: Use DataChecks for partial dependence data validation

Our data checks are mainly used for AutoMLSearch but there might be other places where it makes sense to use their logic to check the data a user passes in.

There is some overlap with this and #968, imo.

Closing criteria:
- A design or draft pr that determines whether it is feasible to use data checks in partial dependence data validation.
- A list of there other places in the codebase that could benefit from leveraging data checks and a recommendation as to whether it's worth using the datachecks there.

Original comment:

> Is it weird not using our own datachecks to accomplish this type of thing? This seems very much like a uniqueness and a no variance datacheck.
>
> _Originally posted by @chukarsten in https://github.com/alteryx/evalml/pull/2120#discussion_r611684152_