fix(spark): preserve aggregate outputs in projections

[bvolpato]

When a Substrait Project follows an Aggregate, Spark conversion folds the projection into the aggregate but replaces the inherited outputs with only the appended expressions and ignores emit mapping. An aggregate row [1, 30] followed by a project adding 99 therefore becomes [99] instead of [1, 30, 99]; an emit selecting the original grouping column also returns the wrong value.

Project.builder()
    .input(aggregate)
    .addExpressions(ExpressionCreator.i32(false, 99))
    .remap(Rel.Remap.of(List.of(0)))
    .build();

Preserve the aggregate's existing named expressions, append the project expressions, and apply emit mapping to that complete list, as required by spec v0.102.0. Keep the existing aggregate folding so projected aggregates retain their Spark round-trip shape.