feat(core): add context to RelVisitor, ExpressionVisitor and FuncArgVisitor

[nielspardon]

BREAKING CHANGE: adds a new context argument to RelVisitor, ExpressionVisitor and FuncArgVisitor

Preparation for #426

Adds the new context argument as a io.substrait.util.EmptyVisitationContext with value null everywhere.

[vbarua]

Had a chance to take a look, and have some high-level notes.

I'm generally a fan of this, but I do worry that it could introduce a lot of noise for visitations that don't require context.

Your usage in a2f0c8e (#426) doesn't strictly require a context parameter like you're introducing, as you could achieve a similar result by just pushing the context before calls to accept, and popping it after. However, that is more error prone and having it as parameter makes it easier to understand the dataflow and makes it harder for users to not propagate the context because they're forced to.

I'm going to ping the Slack to get additional eyes for this, because it does end up being an fairly intrusive API change and I'd like to hear what others think.

[nielspardon]

I'm generally a fan of this, but I do worry that it could introduce a lot of noise for visitations that don't require context.

I also had this concern while working on this. I first had a new set of visitors that I was shipping alongside the existing once. I then decided to just replace the existing once. There are definitely trade-offs for both options.

[nielspardon]

I'm going to ping the Slack to get additional eyes for this, because it does end up being an fairly intrusive API change and I'd like to hear what others think.

make sense. that's why I tagged it as a breaking change since users of substrait-java might be implementing the current visitor and it might be nice to get informed in the release notes that such an intrusive API change has happened.