Pure functions for search result highlighting and snippet extraction.

This namespace provides functions for highlighting matching terms
in search results and extracting relevant snippets.
All functions are pure (no side effects).

Architecture: Functional Core (Pure)

Pure document construction and filter-key conversion for search.

Key functions:
- build-document*       — construct a SearchDocument from field values
- filter-key->json-key  — convert a filter keyword to a snake_case JSON key

The load-time index registry (`defsearch`, `register-search!`, `get-search`,
`list-searches`, `clear-registry!`) is mutable process state and lives in the
shell at boundary.search.shell.registry, keeping this namespace pure.

Pure query-building functions for full-text search.

Two query paths:
- PostgreSQL: to_tsvector/plainto_tsquery + ts_rank + ts_headline
- Fallback (H2/SQLite): LOWER/LIKE-based search

Filter support: callers may pass a :filters map of keyword->string pairs.
Filter values are matched against the JSON "filters" column using
DB-appropriate JSON path syntax.

Pure ranking and scoring functions for search results.

This namespace provides functions for calculating relevance scores,
applying boosts, and normalizing results. All functions are pure.

Architecture: Functional Core (Pure)

Pure Hiccup UI components for the search Admin interface.

All functions are pure — they receive data and return Hiccup structures.
No side effects, no I/O, no logging.

Components:
  field-weight-badge   — weight indicator (A/B/C/D)
  search-result-row    — single result row
  search-results-fragment — HTMX results area
  indices-page         — list all registered indices (full page)
  index-detail-page    — index config + live search test form (full page)