Implement a CacheClient and CacheReader to decrease HTTP requests

[ImJeremyHe]

Rollups always need to query some data frequently. In a scenario where rollups are waiting for L1 finality, they likely keep doing like this:

• given a hotshot transaction hash, fetch the transaction data
• fetch the header regarding the transaction height
• fetch the snapshot height
• fetch the vid common and namespace proof and verify
• fetch merkle proof and verify
• get the finality

In this flow, rollups often have to wait for getting the correct snapshot height. This means rollup keep fetching hotshot transaction data and header again and again until merkle tree proof is verified. These blockchain data are always the same, and rollups actually don't need to fetch them again and again.

Therefore, a client and reader with cache can definitely help rollup decrease the HTTP requests and keep their code clear and straightfoward.