Typed workload binding returns no physical expression for frequency and heap top-k sketches #340

Context

Found by the strict whole-path matrix in ProjectASAP/ASAPQuery-backend#478. Three scenarios fail before backend configuration or ingest because the typed workload planner returns no deployment expression:

1. CountSketch frequency for count_over_time(top_endpoint_qps[10s])
2. CountSketchWithHeap for topk(3, top_endpoint_qps)
3. CountMinSketchWithHeap for topk(3, top_endpoint_qps)

All three stop at the backend test helper's call to control_plane::physical::workload_planner::bind_workload_typed(workload) with:

bind_workload_typed produced a PhysicalExpr

The backend currently pins ASAPPlanner types at revision 5d0b6f6edcac65edc89a72051f37977ab0c83031.

Reproduction

From the ASAPQuery-backend PR branch:

./scripts/e2e.sh whole-matrix

Relevant tests:

• controller_plan_to_query_full_roundtrip_count_sketch
• controller_plan_to_query_full_roundtrip_count_sketch_with_heap_topk
• controller_plan_to_query_full_roundtrip_cms_with_heap_topk

Expected

For each supported workload/query shape, typed binding should produce a physical expression whose three-stage split includes a Backend stage with the selected sketch family and readout:

• frequency/count-over-time -> CountSketch or Count-Min capability
• top-k -> heap-bearing CountSketch/CMS plus a top-k readout

If a shape is intentionally unsupported, return a typed diagnostic identifying the unsupported capability/shape rather than an unclassified None.

Done when

The three strict whole-path tests progress past typed binding and validate the emitted backend/collector configuration. Backend serving failures, if any, should then be reported separately at ingest/query time.