For ExprFunction, replace Environment with EvaluationSession

[dlurton]

I am working on a significant evaluator refactor to work with the new relational algebra. There is a requirement that this be implemented as a refactored copy of the original EvaluatingCompiler to allow customers to switch between the old and new evaluators via runtime configuration. One challenge associated with this is: the new evaluator does not use the same Environment class of the legacy evaluator and (until this PR) and the ExprFunction.call* functions all accepted Environment as a first argument. This PR changes that argument to session: EvaluationSession instead. Unfortunately, this is a breaking API change and customers who have their own ExprFunction implementations. (Although this has gone through a breaking change in the last release anyway and most of our customers probably haven't migrated yet.) There is, however, no other way that I can see to accomplish the goal of exposing the new physical plan evaluator as an alternative to the legacy AST evaluator.

Aside from the above:

• Exposing Environment as public API wasn't a great idea because it exposed implementation details that the customer should never depend on (i.e. all of the properties here except for session).
• No customer that I am aware of uses the env: Environment parameter. Nor is there any reason to that I am aware of, outside of accessing its session: EvaluationSession property.

This PR also makes Environment and several other types internal as well. These other types were all implementation details that customers should not be depending on and could not be made internal previously due to the requirement that Environment be public.

[lziq]

So I think this PR mainly changes Environment from public to internal. I think we should do so, even if this could break users' code, as users should not access it by design.

[lziq]

Nit: Currently the description in kdoc is "the environment for execution". Do you think is makes more sense to change it to "variables environment used during evaluation"?

[dlurton]

I think this version of the class is not just for variables--it includes groups too so the existing wording is more appropriate, IMHO. Anyway, perhaps that is outside the scope of this CR.

[lziq]

Then can we change it as "environment used during evaluation"? I think "execution" in the original definition is not really clear.

[lziq]

Why do we need to add these 2 lines?

[dlurton]

I forgot to remove this--it might be needed a bit later.