https://issues.apache.org/jira/browse/ARROW-12715

P.S. case insensitivity can be achieved by using one of the _upper or _lower kernels, which I think is good enough

We could easily add a flag to toggle insensitivity in RE2 as well. (Would be better than allocating a new string and potentially dealing with Unicode.)

But can PlainSubstringMatcher do case-insensitive matching?

No, but in that case we'd just not dispatch to that path.

What to do about the options class might be awkward. You could subclass MatchSubstringOptions, but I think ultimately we might incorporate an option for case insensitivity into MatchSubstringOptions.

We could add it to MatchSubstringOptions, if we plan to support for all kernels, and leave it unimplemented for now in the non-regex case (it's pretty easy to support in any regex case).

Sounds good. The case-(in)sensitive option should be optional and should default to case-sensitive. We should error in cases where case-insensitive match is specified in the options but is ignored by a kernel or by any path in a kernel. This sounds to me like a separate PR. Should I open a Jira for that?

I filed ARROW-12835. I think this PR should be otherwise good-to-go then, unless we need to also add R bindings.

I think it's good to go. Nothing is needed in the R bindings.

Merged, thanks for the review!