NmtHasher.EmptyRoot has unnecessary double append instead of creating slice with double capacity

If we look at https://github.com/celestiaorg/nmt/blob/29e9433bd79eed40462000626ef987c91b0cf523/hasher.go#L157-L159

we notice this code https://github.com/celestiaorg/nmt/blob/29e9433bd79eed40462000626ef987c91b0cf523/hasher.go#L159

but that code could be made clearer, simpler and even faster (if in hot loop) by

```go
emptyNs2X := bytes.Repeat([]byte{0}, int(n.NamespaceLen)*2)
h := n.baseHasher.Sum(nil)
digest := append(emptyNs2X, h...)
```

Read NmtHasher.EmptyRoot in hasher.go around lines 157-159 and compare the current double-append allocation with the proposed slice construction. Verify the change against the existing test suite and confirm that EmptyRoot preserves its current digest while avoiding the unnecessary append.

Written by the indexing model from the issue text.