Make engine health metrics granular on chunk and result

[mcastorina]

Description:

Some metrics were emitting multiple times per chunk or result, so this adds logic to track it on a per chunk and result basis. This will help us describe the percentage of items getting filtered out in each stage.

Checklist:

• Tests passing (make test-community)?
• Lint passing (make lint this requires golangci-lint)?

---

Note

Low Risk
Metrics-only changes; scanning, detection, and verification behavior are unchanged.

Overview
Aligns chunks_dropped_total with chunks_entered_stage_total so drop reasons are counted once per work item at each stage, instead of firing multiple times inside loops.

At the scanner, no_matching_detectors is emitted only when no decoded variant gets any Aho–Corasick detector match (not once per empty decoded slice). In verification overlap and detect, detector_error is recorded at most once per chunk when any match/FromData call fails, rather than on every failing iteration.

The chunks_dropped_total metric help text now states it uses the same granularity as chunks_entered_stage_total for the same stage, so drop ratios per stage are meaningful.

^{Reviewed by Cursor Bugbot for commit 6927e58. Bugbot is set up for automated code reviews on this repo. Configure here.}

[MuneebUllahKhan222]

LGTM.🚀