Use precomputed hashes for certain padding shares as an optimization

As a celestia-oriented optimization, the following padding shares with

- TxNamespace
- PFBNamespace
- TailPaddingNamespace

Could have the sha256 precomputed. In the `HashLeaf` method, we could simply identify if it is a padding share and if it has one of those namespaces and use the precomputed hash instead. Furthermore we could extend it to the `HashNode` method in the case that there are 2, 4, 8, 16 etc padding shares in a row.

Locate the HashLeaf and HashNode methods and trace how padding shares and the TxNamespace, PFBNamespace, and TailPaddingNamespace are identified. Determine how the existing hash behavior can cover single padding shares and runs of 2, 4, 8, or 16 shares, then verify that the resulting hashes remain correct.

Written by the indexing model from the issue text.