Effect-threaded async-boundary detection for the WasmGC CPS transform (generalise the structural-conservative recogniser) #234

Problem

The WasmGC CPS transform (#225, ADR-013) identifies the async boundary structurally — a let whose RHS calls a recognised async primitive from a hardcoded set (currently http_request_thenable). This was the owner-chosen scope for #225 PR3a (sound, contained, no frontend changes) over the principled alternative.

It is brittle: every new async stdlib primitive must be added to the recognised set, and the transform cannot see a function-defined Async callee at all. It is also a blocker for the fully transparent ADR-013 surface (author writes fetch, backend lowers) since that requires recognising any Async-effecting call, not a name list.

Proposed solution

Thread per-call effect-row information from the typecheck phase into codegen so the async boundary = the first call whose inferred/declared effect row ⊇ Async. This removes the hardcoded set and generalises to user-defined Async functions (also unblocks #225 PR3c Async→Async chaining at full generality).

Notes

• One-way-door-ish: touches the resolve/typecheck→codegen interface (shared infra). Warrants its own ADR.
• Until then the structural-conservative recogniser is the contract; extend the recognised set as a stopgap when new primitives land.
• Surfaced by stdlib Http: typed-wasm target (convergence ABI; shared with Ephapax) #225 PR2/PR3a. Refs stdlib Http: typed-wasm target (convergence ABI; shared with Ephapax) #225.