[Feature] Auto Scaling for TiDB/TiKV in TiDB Operator

[King-Dylan]

Description

Scaling TiDB/TiKV today requires manually updating TidbCluster.spec.replicas. This issue proposes an optional auto-scaling controller that periodically reads metrics (e.g. Prometheus) and automatically adjusts replicas via TiDB Operator.

MVP Requirements

• Triggers: CPU / memory (TiDB); + storage (TiKV)
• Guards: threshold + sustained duration, cool-down
• Safety: min/max/step replica limits, multi-AZ balance, scale-in maintenance window
• Failure handling: retry + timeout, optional rollback, clear events/logs

Phases

• Phase 1: TiDB auto scaling
• Phase 2: TiKV auto scaling (adds storage + data balancing considerations)