Refine batch failure handling

[dnagoda]

📚 Stack (bottom → top)

1. Add batch mode for JSONL function runs #595 Add batch mode for processing multiple inputs via JSONL
2. Preserve JSON input object order #590 Preserve JSON input object order
3. Refine batch failure handling #591 Refine batch failure handling
4. Reuse the compiled provider across runs #592 Reuse the compiled provider across runs
5. Compute humanized bytes lazily #593 Compute humanized bytes lazily
6. Add minimal batch output with --batch-full-output opt-in #594 Add minimal batch output with --batch-full-output opt-in

Stacked PRs — review bottom-up.

---

Why

The initial batch mode counted every run that returned Ok as a success — even
when the function itself returned success: false — so batch summaries were
wrong.

What

• Keep --batch-continue-on-error (opt-in; batch stops on the first error by
  default). No new flag.
• Count real successes/failures from FunctionRunResult.success; track
  processed / successful / failed.
• Treat a function-level failure (success: false) like an error: it stops the
  batch by default, and is recorded-but-skipped under --batch-continue-on-error.
• The stderr summary reports processed, successful, and failed counts.
• Switch integration tests to assert_cmd::cargo::cargo_bin! (the
  non-deprecated API) in place of Command::cargo_bin.

Testing

• cargo test batch_
• cargo test