Implement caching layers for whurl client

[nicmostert]

Overview

Implement a robust, layered caching strategy for the whurl client, with clear separation of responsibilities and interfaces for each layer.

Caching layers (sub-issues):

1. Domain-layer cache (canonical data):
   • Caches raw, unfiltered site and measurement lists in memory (and optionally disk).
   • API examples: client.list_sites(refresh=False), client.check_for_site(name, refresh=False).
   • Handles normalization, index building, refresh logic.
   • First to implement (see sub-issue).
2. HTTP-layer cache (transport-level cache):
   • Intercepts GET/HEAD requests, uses ETag/Last-Modified and Cache-Control to serve fresh or revalidated responses.
   • Respects server cache headers and conditional requests.
   • Pluggable backend (memory/disk), with a size-limited LRU.
   • Sub-issue details to follow.
3. Query memoization layer (hot filtered queries):
   • Small, short-lived cache for expensive filtered queries (e.g., measurement filters).
   • Built atop domain-layer cache; optional, not persisted.
   • Sub-issue details to follow.

Each layer should be modular, well-documented, and testable. Sub-issues will cover detailed API, implementation, and acceptance criteria for each layer.