Issue Description:

Background:
The current py-libp2p implementation suffers from multiple isolation issues preventing it from seamlessly communicating with the broader IPFS network (specifically Kubo nodes).

Identified Problems:

1. Incompatible CID Generation: MerkleDag uses CODEC_RAW for leaf chunks instead of dag-pb + UnixFS, producing CIDs entirely foreign to Kubo.
2. Flat DAGs: The DAG structure creates flat, oversized root blocks for large files instead of a balanced layout, breaking Kubo's 174-link limits.
3. Ephemeral Storage: MemoryBlockStore loses all fetched blocks on exit, making the node impractical for long-running environments.
4. Kubo Disconnects: Fetching many blocks sequentially opens hundreds of streams, causing Kubo to issue GO_AWAY disconnects.
5. DHT Record Rejection: DHT value records lack proper cryptographic signing, causing them to be rejected by Kubo peers.

Proposed Solution:
This epic will overhaul the Bitswap and Kademlia DHT sub-protocols to achieve the following:

• UnixFS / Balanced Layout: Implement UnixFS Data wrapping for leaves and a balanced.Layout tree builder to generate byte-for-byte identical CIDs to Kubo.
• Canonical Serialization: Enforce Links-before-Data ordering for DAG-PB protobufs.
• Batch Message Streaming: Implement get_blocks_batch() to group wantlists and reuse streams.
• Persistent Storage: Implement FilesystemBlockStore and a BlockService caching layer.
• Payment Extension (1.3.0): Create a modular IBitswapExtension architecture, introducing BitswapPaymentClient, PaymentLedger, and BlockPricingEngine.
• DHT Signing: Implement signed/verified Peer and Value records in the Kademlia DHT.
• Type Safety & Streaming: Replace magic integers with typed Wantlist classes and introduce io.IOBase streaming for large file ingestion.