This is a rough approximation: not every hash is counted.
Furthermore, there can be considerable overhead in moving
data around.
If the generic hash is selected, this counts
hash calls. If a specific hash is selected, then the
number of internal blocks/permutations is counted.
Assumes for XMSS that the state after digesting the public
seed is precomputed.
Assumes the worst case that signatures are all zeroes. In the average
case fewer hash calls will be required.
Will be slightly larger for an actual scheme:
does not include the signature sequence number, algorithm
specification and digest randomization value.
We take w to be the length of the chains
as in XMSSMT and SPHINCS+.
What is called w in LMS/HSS
corresponds to 2w.
In LMS/HSS the size of the hashes in the OTS, called n
can be different from the size of the hashes in the Merkle tree
denoted by m. We assume n=m.