Correct the docs on when constructing a Digest can fail

[SeanTAllen]

PR #92 made the Digest constructors partial, so construction raises when OpenSSL cannot allocate the context. Three statements were left describing the old behavior.

ssl/crypto/crypto.pony said constructing a Digest cannot fail and that the failure surfaces at the first append or final, and it omitted the constructors from the list of partial functions.

.release-notes/next-release.md had two wrong clauses about the same thing. One said a digest that could not be created reports it at the first append or final; the other said construction now raises rather than returning a digest that fails on every later call. Before #92 a failed construction crashed, so both described a state that never shipped.

Closes #106

[SeanTAllen]

Autonomous-mode decision log — PR-specific

No release note. The wrong docs arrived with the still-unreleased #92, so nothing incorrect shipped to users; correcting them before release needs no note.

Scope grew after the pre-PR review. I first fixed only the two statements #106 quoted. The reviewer found a third: a sibling note ("Make Digest and HmacSha256 raise on failure") said construction now raises "rather than returning a digest that fails on every later call." I checked the released 2.1.0 constructor — it was non-partial and dereferenced a possibly-null context, so a failed construction crashed. Both wrong clauses describe a state that never shipped, so I corrected both to say "rather than crashing," matching the code and each other.