Mixed missing and non-missing mutation times clarification

[jeromekelleher]

#726 added a requirement that we cannot mix known and unknown mutation times. However, I think this is a bit too strict, because we only need to insist that known and unknown mutation times can't be mixed within a site to resolve the ordering problems.

Should we relax to the checking we don't mix known and unknown times within a site? We could bundle this into the 0.99.4 C API release we need to do anyway (#747).

cc @petrelharp, @mufernando

[petrelharp]

Yes, definately. That'd make it significantly less likely this restriction would affect anyone.

[benjeffery]

Ok, sounds sensible.