Breadcrumbs aren't read individually by NVDA

[adamwoodnz]

As explained by @alexstine

This is a Windows-specific bug. Voiceover and Windows screen readers work differently in the way that Voiceover traverses the content one item at a time where NVDA reads the links like a sentence due to the display: inline; set. I think that is okay for breadcrumbs, but to get the best verbosity results, it needs to be "Link 1, Link2, etc." Breadcrumbs can be tricky... I wonder what would happen if we tried rendering them as display: block;. Then there likely would not be a need for the screen reader text anymore.

We should try using display: block breadcrumbs within a flex row container if the attempted fix in #1641 doesn't work.