PR 1 — Deep aggregate unwrapping (standards-aligned, low risk)

Goal: Fix PE056 cases where QUERY source is a named aggregate type or SELECT-of-named-aggregates by doing deep type unwrapping rather than shallow TYPEis_aggregate.
Agent input

Implement deep aggregate detection + element-type extraction used by the QUERY resolver and by TYPE_retrieve_aggregate():

Add a helper:

Type TYPEget_aggregate_base(Type t) (or better name: TYPEaggregate_base), exported in include/express/type.h, implemented in src/express/type.c.
It should unwrap through defined type chains until it hits an aggregate constructor (set_, bag_, list_, array_, aggregate_), and then return that aggregate’s element type (tb->base).
It must return NULL if t does not resolve to an aggregate.

Update QUERY resolution in src/express/resolve.c:

Replace shallow TYPEis_aggregate(expr->return_type) checks with TYPEinherits_from(expr->return_type, aggregate_).
When it’s an aggregate (possibly through typedefs), use TYPEget_aggregate_base() to derive the query variable type.

Update TYPE_retrieve_aggregate():

When iterating SELECT members, use TYPEinherits_from(member, aggregate_) and TYPEget_aggregate_base(member) rather than shallow TYPEis_aggregate and direct ->body->base.
Keep the existing “common element type” check: if different member aggregates have different element base types, return NULL.

Tests:

Add a small .exp schema under test/ that covers:
    a defined type TYPE my_set = SET OF INTEGER;
    a SELECT whose members are my_set and another defined aggregate (same element type)
    QUERY over a variable of the defined aggregate type
    QUERY over a variable of the SELECT-of-aggregates type (only if all SELECT members are aggregates; if you keep it strict, reflect that)
Ensure the tests verify pass behavior only (don’t add “must fail” tests unless the harness supports negative tests).

Non-goals in PR 1:

Do not implement flow-sensitive narrowing.
Do not implement TREAT syntax yet.
Do not change global typing rules for SELECT beyond the deep-unwrapping.

Acceptance criteria:

Existing behavior unchanged for inline aggregates.
Named aggregate QUERY sources work.
Any existing “buggy expr-src-aggregate.exp” test is moved into passing tests and passes.
AP242 may still fail on styled_item WR3 (expected); that’s for later PRs.