feat(storage): introduce checkpoint ring buffer storage experiment

Overview

The current memory storage implementation uses Go synchronization primitives to protect workflow state.

This issue explores alternative storage layouts optimized for high-frequency checkpoint operations.

Proposed Work

Implement an experimental ring buffer based storage component:

• Create a fixed-size checkpoint buffer
• Manage read/write positions safely
• Evaluate atomic operations using sync/atomic
• Add benchmarks comparing current storage and ring buffer implementation

Expected Outcome

Provide benchmark results for:

• Write throughput
• Read performance
• Memory allocation behavior
• Concurrent access patterns

Requirements

• Existing storage interfaces must remain compatible
• Current MemoryStore behavior should continue working
• New implementation should be isolated and testable

Notes

This is a performance exploration. The default storage backend should remain unchanged until benchmarks justify adoption.

Start by locating the existing MemoryStore implementation and storage interfaces, then review how checkpoint reads and writes are synchronized. Design the isolated fixed-size ring buffer around compatible interfaces and safe read/write positions. Add tests and benchmarks covering throughput, reads, allocations, and concurrent access, while confirming the current MemoryStore behavior remains unchanged.