See #5648 (comment)

We're currently hashing all the blob commitments in the mempool for each request. Instead, we should store the blobs keyed by versioned hash inside the mempool, to retrieve them inexpensively during engine_getBlobsV2.