Add support for multi-layer caching (memory → Redis fallback)

Description

Currently, Loco supports configuring either an in-memory (Mako) cache or a Redis cache for application-level caching. However, many production environments would benefit from a multi-layer cache system — where the application checks memory first and falls back to Redis if a key is not found.

This would combine the speed of in-memory caching with the persistence and distribution benefits of Redis.

Proposed behavior

Add support for a two-layer cache chain, such that:

1. The cache lookup first checks the in-memory (Mako) cache.
2. If the key is not found, it falls back to Redis.
3. When setting a value, it updates both layers.

Optionally, developers should be able to configure:

• TTL sync between layers.
• Whether both layers are enabled (e.g. only memory, only Redis, or both).
• Error handling if one layer is unavailable.

Example configuration

cache:
  kind: Multi
  layers:
    - kind: Memory
      max_size: 100
    - kind: Redis
      uri: "redis://localhost:6379"
      max_size: 10

Benefits

• Faster cache hits for hot keys via memory.
• Reduced Redis load.
• Persistent distributed cache for scale-out deployments.

References

• Loco cache documentation
• Inspiration: multi-tier caching in frameworks like Spring, Django, and ASP.NET.