docs: outline how ASAPPlanner supports a new end-to-end use case

[zzylol]

Why

planner-layering.md designs the planning stages, but there is no guide for bringing a new use case onto ASAPPlanner: how to describe its workloads, which ASAP Primitives fit, what to add when a primitive is missing, and how a deployment runtime integrates with the planner so the use case shows an end-to-end benefit.

What

Adds one file, docs/design_docs/proposals/asapplanner-e2e-use-case.md (111 lines). No other doc or code changes. It is an outline: each section holds key points marked "to be expanded".

1. Defining the query and data workloads from recorded traces. The output is a PlanningWorkload: per query (text and language, accuracy target, latency, recurrence and predictability, time selection) and per source (schema, arrival and retention, cardinality and value distribution). It also covers the overlap analysis that makes ASAP worthwhile and declaring unmeasurable statistics as unknown.
2. Query patterns → ASAP Primitives: a table from intent (count/sum, quantile, top-k, distinct count, frequency moments, high cardinality, overlapping windows, shared sub-expressions) to candidate primitives and the workload conditions each needs. It also states the eligible / valid / selected distinction.
3. Adding a new ASAP Primitive: eight steps (capability declaration, Pass 1 offer, sizing and error bound, Stage 2 operator, Stage 3 pricing, executor kernel over the sketch library, sketch-bench measurements, tests). This extends planner-layering.md's "Adding a new summary family".
4. Integrating with a deployment: a six-step workflow (collect → plan → validate and install → execute → observe and re-plan → compare with the exact baseline) and a table of interfaces between deployment and planner (e2e_plan, PlanningModels, capabilities, PhysicalASAPDAGDocument, the selection report, feedback).

How

Written as a design proposal next to planner-layering.md. It links that doc and asap-primitive-schema.md and keeps stage names and terms aligned with them.

Key code interfaces added / modified

None (documentation only)

Before this PR

The stage design existed, but no doc said how to go from a use case to a planned, deployed workload, or how to add a missing primitive end to end.

After this PR

There is one outline that covers that path, section by section, ready to be filled in.

Verification

• Doc-only. Build, tests, screenshots and measurements: not applicable.
• Links: planner-layering.md, its "Adding a new summary family" section, and asap-primitive-schema.md exist on main.

Limitations and follow-ups

• Outline only. The Contents list still carries [TODO: just fill], and every section says "to be expanded".
• Open inline review question: should the accuracy model be a Stage 1 input, per planner-layering.md?
• The deployment-capabilities interface and a single bundle for all deployment inputs are not defined yet.
• Tracking issue: Tracking: end-to-end use case support for ASAPPlanner (inputs, missing components, adding summaries) #685.

Human review — do not complete with an agent

• The MVP boundary is correct.
• New conceptual layers or public interfaces are necessary.
• The before/after description matches the intended product behavior.
• Human reviewer:
• Decision and rationale: