feat: clarify the meaning of plans

[westonpace]

#612 and #613 highlighted a number of ambiguities in plan interpretation. I have made an attempt to clarify these based on my interpretation.

[ingomueller-net]

This is very helpful -- thank you, @westonpace!

I think the new text makes a good job in distinguishing root and non-root relations and what their purposes are. The only comment I have is that the current terminology is not in line with the field names in the protobuf definitions: there we have one "plan" (i.e., one Plan message) that contains several (root or non-root) "relations" (i.e., several PlanRel messages), whereas here we have several "plans" that contain "a tree of relations" (which could be read as "one tree of relations").

How about aligning the terminology with the one from protobuf? Something like "Substrait is designed to allow users to express [...] transformations. [...] Transformations are expressed in a 'plan' that consists of several 'relations'. Each relation is a tree of 'relational operators'. We distinguish between 'root relations', which are interpreted as the possible result(s) of the plan depending on the context, and 'non-root relations', which represent intermediate results that other relations can refer to but which by themselves are not part of the semantics of the plan. (In particular, they can be removed, if unused.)"

If you agree, I can make the changes, which look quite mechanic, but feel free to do them yourself if you don't want to wait for the round-trip.

[jacques-n]

This looks really good. Seems like a few small edits proposed that look good and then we should be able to merge. Ping @westonpace to push over the finishline.

[westonpace]

Addressed comments. @jacques-n

[jacques-n]

I think this is a good set of clarifications. Going to merge.