feat: add context enrichment scaffold for weather and jurisdiction fields

[Harshvardhan-91]

Summary

Fixes: #484
Adds a provider-agnostic context enrichment scaffold for enriching incident reports with trusted weather and geographic context.

This PR introduces a small backend architecture layer that can later support providers such as NOAA, OpenStreetMap, Nominatim, or local department boundary datasets.

The goal is to avoid asking the LLM to infer objective facts such as weather, coordinates, jurisdiction, or district when those values can eventually come from deterministic trusted sources.

Why

FireForm’s core pipeline currently depends heavily on narrative extraction from text/voice input.

For many emergency and wildfire reports, some fields are factual environmental/geographic context rather than narrative facts. Examples include:

• wind speed,
• humidity,
• temperature,
• latitude / longitude,
• jurisdiction,
• operational district.

These values are important for incident reporting, but they should not be guessed by an LLM.

This PR creates the foundation for a future enrichment step where:

incident text / voice transcript
        ↓
LLM extracts narrative fields
        ↓
trusted context enrichment
        ↓
weather + geography providers
        ↓
enriched incident JSON
        ↓
template mapping + PDF filling

Future scope

This could later be extended into a unified quality scoring layer across extraction, validation, and enrichment signals