This document source holds on to a TransformerInterface which knows how to serialize itself. The issue is that the implementations of TransformerInterface hold on to projection executors which represent the projection after optimization and lowering from the projection AST, which is later discarded.

The query shape spec dictates that the query shape "should be calculated by serializing the first parsed query representation we have. For example, a MatchExpression, a projection_ast::Projection, or a Pipeline."

Therefore, this ticket will need to modify the TransformerInterface to additionally hold on to a `projection_ast::Projection` so that it can serialize that structure instead in the case that the serialization options indicate that the query shape should be shapified/redacted.