[gr8man]

Description
In both RedisHandler::save() and PredisHandler::save(), saving a cache item with a TTL previously executed two sequential network round trips:

1. hMset (hmset)
2. expireAt (expireat)

This PR executes both commands in a single network round trip using Redis pipelining ($this->redis->multi(Redis::PIPELINE) in RedisHandler and $this->redis->pipeline(...) in PredisHandler).

Behavior & Compatibility Analysis

• Data Structure: Preserves the existing Hash structure (['__ci_type' => ..., '__ci_value' => ...]). This ensures 100% compatibility with existing cache keys, hMget(), hIncrBy() (increment()/decrement()), and getMetaData().
• Permanent TTL ($ttl === 0): Continues to execute a single hMset without expiration.
• Edge-case / Error path consideration: In normal usage, BC is completely preserved. The only semantic difference is on error paths: if HMSET fails (e.g. if an existing key holds a non-hash data type), EXPIREAT in the pipeline batch will still execute on that existing key, while save() returns false as expected (previously expireAt was not reached if hMset failed). Using a Lua script to conditionally expire was considered, but avoided to prevent cluster and proxy portability complexity.

Benchmark (5,000 save() operations, local loopback)

• Sequential (2 round trips): 528.25 ms (~0.106 ms per save)
• Pipelined (1 round trip): 298.33 ms (~0.060 ms per save)
• Local speedup: 1.77x (43.5% latency reduction). Over network links (e.g. cloud/separate container with 0.5–2 ms RTT), this eliminates an entire network round trip per save (~50% network latency reduction).

Checklist:

• Securely signed commits
• Component(s) with PHPDoc blocks, only if necessary or adds value (without duplication)
• Unit testing, with >80% coverage
• User guide updated
• Conforms to style guide