Support TaskFlow call syntax on stub tasks for the Go SDK

[jason810496]

Why

@task.stub tasks can only be declared argless today, so a Go task that needs an upstream's output has to hand-write GetXCom calls (with the upstream task_id hard-coded in Go, duplicating the wiring the Dag file already expresses). This PR makes the natural TaskFlow call work across the language boundary:

@task.stub(queue="golang")
def transform(country: str, extracted: dict): ...

with DAG(...):
    transform("uk", extract())   # extract() is a normal Python @task

// The runtime binds "uk" onto country and pulls extract's XCom into extracted.
func transform(ctx sdk.TIRunContext, log *slog.Logger, country string, extracted map[string]any) error

What

• Parse time (providers/standard): _StubOperator binds the TaskFlow call to the stub's signature and serializes an ordered positional-arg spec with the Dag (_arg_bindings): each entry is an XCom reference or a JSON literal, with data_type derived from the parameter annotation.
• Server (execution API): ti_run returns the spec as a new optional TIRunContext.arg_bindings field — extracted from the serialized-dag blob only for _StubOperator tasks, and stripped for older API versions (Cadwyn change in the in-progress 2026-06-30 block). The supervisor needs zero changes because TIRunContext is embedded verbatim in StartupDetails.
• Supervisor schema: new in-progress version 2026-07-30 (AddArgBindingsToTIRunContext); the downgrade path strips the field, so bundles pinned to 2026-06-16 keep working unchanged.
• Go SDK: new pkg/binding package. Analyze classifies parameters once at registration (injectables like context.Context/*slog.Logger vs JSON-decodable data parameters); Resolve binds the spec onto the data parameters in order — literals decode directly, XCom entries pull on demand. Arity or declared-type mismatches fail the task loudly before the body runs, replacing the old silent reflect.Zero fill.
• Fail-loud v1 exclusions at parse time: .expand()/.partial() (via a new generic supports_expand opt-out), map/zip/concat XComArgs, *args/**kwargs, context-key parameter names, and non-JSON-serializable literals.

---

Was generative AI tooling used to co-author this PR?

• Yes, with help of Claude Code (Fable 5) following the guidelines