docs: document inspecting and editing batch slices from the console

[reidmorrison]

Summary

Expands the Batch Guide's console/debugging coverage. Previously the only documented way to inspect a batch job's slices was a single each_failed_record example. This adds two sections, all verified against lib/rocket_job/sliced/ and lib/rocket_job/batch/:

• Inspecting slices from the console — counting slices (job.input.count, per-state .failed.count, etc.), fetching/iterating via the state scopes (queued/running/failed/completed) plus first/last/each/where, reading records and per-slice metadata (first_record_number, failure_count, failed_record, ...), inspecting the embedded JobException, and the previously-undocumented group_exceptions aggregation helper. Notes that it all applies to job.output and named categories.
• Editing and retrying failed records — correcting a bad record in place (records is a plain Array; save! persists it) and retrying, including the resume-from-processing_record_number behavior and the key distinction: slice.retry! works while the job is still running, but a failed job needs job.retry! since workers only process slices for a running job.

Notes

• Docs-only change (docs/batch.md), no code changes.
• Follows existing kramdown conventions on the page: ~~~ruby tilde fences and bold-prose "Example:" labels so they stay out of the auto-generated TOC (the two new ## headings do appear in the TOC).

🤖 Generated with Claude Code