Allow Time::Span.new with positional + nanoseconds

[bcardiff]

Fixes #8825

[straight-shoota]

I don't follow why we need this to fix #8825. What's wrong with my suggestion from #8825 (comment) ?

I suppose we could still follow with this because nanoseconds is more likely to be passed as named parameter than other arguments. I'm not sure, but it might have even been a named-only argument at one point.

[bcardiff]

The idea was to use named argument for subseconds units since the nano/micro was not as obvious as the rest of the parts.

Some db drivers use the named argument so without this we are breaking that code without a deprecation cycle. People are forced to update in order to use nightly rather that noticing a deprecation warning on nightly.