feat(core): implement DiskCache using sled

[JuliobaCR]

Description

Adds DiskCache, a persistent, disk-backed cache implementing the CacheProvider trait (#412) using sled, a high-performance embedded database. This allows large or long-lived data to be stored on the local filesystem and survive application restarts, without a custom eviction/LRU scheme.

How It Was Done

• crates/core/src/cache/disk.rs: DiskCache struct wrapping an Arc<sled::Db>, with new(path) and default_location() constructors (the latter reuses the same directories::ProjectDirs convention as CacheStore).
• Implements CacheProvider::{get, put, remove, clear}: values are serialized/deserialized with bincode before hitting the sled tree; keys stay &str, consistent with the trait.
• Relies on sled's internal lock-free concurrency guarantees for safe concurrent access — no external mutex needed. Blocking sled calls are offloaded via tokio::task::spawn_blocking so the async executor is never blocked.
• Serialization/deserialization failures map to the typed GratError::CacheSerializationError / CacheDeserializationError variants added in feat(core): implement CacheProvider trait #412; lower-level backend I/O failures (open/get/insert/remove/clear, task join errors) reuse the existing GratError::CacheError(String) variant, consistent with CacheStore.
• Added sled = "0.34" as a workspace dependency, following the same { workspace = true } pattern as redb.
• Added a flush() method (via sled's async flush_async) so callers needing durability guarantees (e.g. before a restart) can await it explicitly, since sled lazily batches writes.

Issues Encountered (If Any)

• bincode is not self-describing, so deserializing into the wrong type doesn't reliably error unless the stored bytes are too short for the target type's layout — the type-mismatch test exploits that instead of relying on bincode detecting the mismatch directly.

Related Issue

Closes #403

How It Was Tested

cargo test -p grat-core --lib cache::disk — 11/11 passing: put/get roundtrip, put overwrites existing entry, cache miss returns Ok(None), remove (present and missing key), clear, persistence across a Db close/reopen cycle (via explicit flush()), concurrent put/get from 20 tokio tasks, and a typed deserialization-error case.

cargo clippy -p grat-core --lib --tests -- -D warnings and cargo fmt --all -- --check are clean for the new code.

Screenshots / Video (If Applicable)

N/A — no UI impact.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: f717c2af-fd47-40f8-b52a-0a554a6a28cd

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[codeZe-us]

Pr reviewed