Regression-breaking PRs should be able to supply an "override" value so tests pass

[rcabell]

When submitting an answer-changing (or metadata-changing) PR, regression tests fail as expected, but in cases where the change is intentional, there should be a way to indicate the new, correct answer independently of the updated code.

Expected Behavior

Regression Tests should pass if code changes produce a different, but expected answer.

Current Behavior

Regression Tests fail due to changed values, even if changed values are deemed correct externally.

Possible Solution

Provide a file with expected new results, which also creates traceable documentation of the evolution of changes

Related Issues

#50 Testing: checks for data and metadata should be separated