Signature digest calculation varies

AFAICT

• (At least some) ED signatures eg router-sig-ed25519 have a fixed string prefix included in their digest but RSA signatures do not. netdocs are not readily confusible for other protocol items, but it is fine to have a specific prefix. But, IMO, there should be one fixed prefix which should be used for all netdoc signatures - after all a netdoc's type is evident from its introductory keyword.

• ED signatures eg router-sig-ed25519 stop after the after-keyword space; RSA signatures eg router-signature include the item line arguments (of which there are none specified) and the newline. I can see no reason to include the signature item keyword, and including the optional arguments is a hazard to extensibility. The digest should stop at the start of the first signature item. That would make the parsing uniform.

• (Also there should be a standard way to identify a signature item so that future code can skip over unknown ones and not include them in digests.)