feat(studio): add scheduled runtime tasks

[evanlowe]

Summary

• split scheduled-task processing into stateless minute scanning, durable ready-queue dispatch, independent Runtime execution, and terminal result persistence
• persist user-scoped definitions, minute due indexes, ready work, run history, and compare-and-set job locks in TOS
• deploy separate scanner and worker functions for both Volcengine and BytePlus; scanner completion is independent of Runtime duration
• prevent concurrent execution of the same task, preserve ready work across transient failures, reject late terminal overwrites, and retain full sanitized Runtime errors for retry
• use Apps SDK UI controls, add expandable final answers, simplify loading, and place Automation below Scheduled Tasks without the Beta badge
• package the updated Studio WebUI assets

Reliability validation

• 500 tasks due in the same minute: 500 pointers scanned, 500 Runtime executions, 500 terminal successes, and 0 ready entries remaining
• scan completed before Runtime execution; scan 436.509 ms, total 906.737 ms
• execution latency: p50 239.060 ms, p95 435.928 ms, p99 452.785 ms, max 457.160 ms
• queue latency: p50 443.387 ms, p95 739.910 ms, p99 814.360 ms, max 834.926 ms
• local real Runtime E2E: queued -> running -> success, persisted Runtime session and final answer
• BytePlus cloud real Runtime E2E: successful execution, complete Runtime error propagation, retry through the next scanner cycle, new Runtime session, and terminal success
• deployment tests cover both Volcengine and BytePlus scanner/worker configurations

UI validation

• desktop 1440x900: loading, list, detail, running, success, failure, and retry states
• mobile 390x844: detail, collapsed final answer, expanded final answer, and collapse action
• verified alignment, spacing, typography, hierarchy, clipping, overflow, responsive behavior, sidebar order, and Beta badge removal

Automated validation

• frontend tests: 752 passed
• Python scheduler, benchmark, and Studio deploy/update tests: 164 passed
• production WebUI build and packaged asset verification: 81 files and 226 internal references
• Ruff: passed
• Pyright: 0 errors and 0 warnings
• pre-commit Ruff formatting/checks and hardcoded-secret scan: passed