Proposal: Do not require else branch on valued while with comptime-known true clause

[ghost]

Currently, any valued while requires an else branch, even if that branch can be statically known to never evaluate. The sensible thing would be else unreachable, but say the code is refactored, and suddenly the clause is runtime-known -- then there's a new point of failure, either a crash if caught in testing or UB if not.

After discussion with @ifreund and @SpexGuy (by which I mean, they discussed it -- I'm just the messenger), we believe the best solution is to allow leaving off the else branch in this case. That way, you don't have to write any dead code, and if a refactor makes it live, you get an error instead of a crash or UB.

At first blush it may seem strange to special-case comptime clauses like this, but in fact there is precedent -- statically-known untaken if and switch branches are not analysed, for instance, which is crucial in many areas. One thing which was considered to make such a thing seem more normal was allowing the clause to be optional, and renaming while to fit, however I personally believe this to be a bad idea; see #7826 for reasoning.

[Vexu]

Duplicate of #707