What is the problem the feature request solves?

In CometCastSuite we have a shared method castTest which executes queries against Spark and Comet and compares results.

In the case where we expect the cast to fail, we compare error messages.

When running against Spark 3.4 we expect the error messages to be identical, but for 3.2 and 3.3 we have been adding custom handling for different types of errors, and this area of code is becoming a bit hacky (my bad) and I think we need a better mechanism.

I propose adding a capability for individual tests to provide their own lambda functions for checking for expected errors. I plan on creating a PR with a proposal.

Describe the potential solution

No response

Additional context

No response