Implement HTMX-based web application for collectables inventory management

[Copilot]

Builds out the full web application for invdex — a stateless collectables inventory system with LLM-assisted cataloging, backed by PostgreSQL and configurable object storage.

Architecture

• internal/config/ — Env-based configuration (DB, storage backend, LLM endpoint)
• internal/database/ — PostgreSQL with auto-migrations, CRUD for items/tags, search with proper LIKE escaping
• internal/storage/ — Abstraction over S3-compatible and GCS backends (via S3 interop), presigned URLs for serving
• internal/llm/ — OpenAI-compatible chat completions client; generates item suggestions and powers conversational assistant
• internal/handlers/ — HTTP handlers with embedded HTML templates, HTMX for partial page updates

UI

• Item grid with tag-based filtering sidebar and live search (keyup debounced via HTMX)
• Add/edit forms with image upload (multipart → object storage) and "✨ Suggest" button for LLM-generated metadata
• Per-item AI chat panel for identification, valuation, and categorization assistance
• Tag management CRUD

Key decisions

• Standard library net/http with Go 1.22+ method routing — no external router dependency
• Templates embedded via //go:embed for single-binary deployment
• Storage key generated server-side (UUID-based path), HTML-escaped before rendering to prevent XSS
• HTMX loaded with SRI integrity attribute
• chatWithLLM delegates to the same OpenAI-compatible endpoint used for suggestions, providing agent-like conversational ability without additional infrastructure

// Storage is configurable via STORAGE_BACKEND env var
store, err := storage.New(ctx, cfg) // returns S3Storage or GCSStorage