feat: PRISMA flow report generation

[BenjaminNavet]

Description

After each pipeline run, automatically emit a prisma_report.md documenting the full provenance of papers: records identified per source API, records after deduplication, records after quality filtering, and final top-N count. Optionally include a Mermaid flowchart block for direct embedding in manuscripts.

Justification

PRISMA 2020 (Page et al., BMJ 2021, doi:10.1136/bmj.n71) is the standard reporting checklist for systematic reviews, endorsed by nearly 200 journals and organisations and cited in over 60,000 systematic review reports. Its core artifact — the flow diagram — must document exactly how records were identified, deduplicated, and excluded. SciLEx performs all these steps mechanically but currently generates no audit trail. All numbers are already computed during the pipeline run; this feature is purely additive logging.

Affected files

• new scilex/reporting.py
• scilex/aggregate_collect.py
• scilex/run_collection.py