HTTP routes for the admin interface.

This namespace wires the admin panel's normalized web routes to their
handlers. Handlers live in `boundary.admin.shell.http.handlers.*` and their
shared plumbing (middleware, error mappings, query/form parsing, handler
helpers) lives in `boundary.admin.shell.http.support`.

The admin UI provides:
- Entity list pages with search, sort, and pagination
- Entity detail/edit pages with forms
- Create and update handlers with validation
- Delete and bulk delete operations
- HTMX fragment handlers for dynamic updates

All routes require authentication and admin role.
Routes follow normalized format for consistent interceptor application.

Create / update entity handlers.

Delete + bulk-delete entity handlers.

Entity detail / edit page handlers (view existing + new-entity form).

Inline editing handlers (Week 2): per-field edit widget, update, cancel.

Admin home + entity list handlers.

Renders the dashboard, the entity list page, and the HTMX table fragment.

Shared plumbing for the admin HTTP layer.

Leaf namespace requiring no handler or route namespaces. Provides the
error mappings, query/form parsing, and handler helpers used by the handler
namespaces and the route definitions in `boundary.admin.shell.http`.

Integrant lifecycle management for admin module.

This namespace defines Integrant initialization and shutdown methods
for all admin module components:
- Schema provider (ISchemaProvider implementation)
- Admin service (IAdminService implementation)
- Admin routes (HTTP routes)

Dependencies are injected via Integrant refs, maintaining FC/IS separation
and enabling testability with mock implementations.

Permission assertions for admin access control.

These wrappers throw typed `ex-info` (`:type :forbidden`) when a permission
check fails, so they live in the shell — the pure decision logic (the
`can-*?` predicates and `explain-admin-access-denial`) stays in
`boundary.admin.core.permissions`. HTTP handlers call these guards; the
platform error interceptor maps `:forbidden` to a 403 response.

Schema repository implementation for database metadata retrieval.

This namespace provides the concrete implementation of ISchemaProvider,
connecting to the database through existing adapter protocols to fetch
table and column metadata.

Responsibilities:
- Fetch raw database metadata using adapter protocols
- Coordinate with pure core functions for parsing and merging
- Cache entity configurations for performance
- Provide entity discovery based on configuration

Caching: computed entity configurations are cached in an atom inside the
long-lived SchemaRepository component (populated on first access). Table
schemas only change at migration/deploy time, so the cache lives as long
as the component. Use `reset-cache!` (or recreate the component via
(ig-repl/reset)) to invalidate after a migration.

Admin service implementation for CRUD operations on entities.

This service provides database-agnostic CRUD operations on any entity
managed by the admin interface. It coordinates between schema providers,
permission checks, and database operations.

Responsibilities:
- Execute CRUD operations with observability
- Apply pagination, filtering, sorting
- Validate permissions and entity access
- Transform data between DB and application formats
- Handle soft/hard deletes based on schema