fix: fall back from api_server deliver=origin to home channels (#69304)

[aurorabotticus-svg]

Fixes #69304

Problem

When a cron job is created from an api_server session with deliver=origin, the origin resolves to {platform: "api_server", chat_id: ...}. But api_server is request/response only — it has no send() adapter. The job runs and succeeds (last_status: ok) but delivery silently fails with API server uses HTTP request/response, not send().

This is invisible to the creator — they scheduled a job and nothing ever comes back.

Fix

In _resolve_single_delivery_target(), detect when deliver=origin resolves to api_server platform and log a clear warning, then fall through to the home-channel fallback (which already exists for jobs with no origin at all):

if origin.get("platform") == "api_server":
    logger.info(
        "Job '%s' has deliver=origin from api_server session; "
        "api_server cannot deliver — falling back to home channels",
        job.get("name", job.get("id", "?")),
    )
else:
    return { ... }  # normal origin delivery

The fallback path (for platform_name in _iter_home_target_platforms()) then picks up a deliverable platform.

Verification

• py_compile passes
• Jobs from non-api_server origins continue to work unchanged
• Jobs from api_server sessions now deliver to a home channel instead of silently failing

[alt-glitch]

This was generated by AI during triage.

Related: #69350 preserves origin delivery through a request-scoped delivery-platform hint, while #69384 adds generalized undeliverable-origin validation and creation-time notices. This PR applies only the api_server home-channel fallback; a maintainer should choose the policy.

[aurorabotticus-svg]

Closing — duplicate of existing PR. Apologies for the noise.