Review Draft integrity

[annevk]

There's a minor concern when updating a Review Draft of a standard that has additional sources besides the Bikeshed source (HTML's version of this is tracked over at whatwg/html-build#154).

If you publish a RD, then modify the non-Bikeshed source, and then modify the published RD's source, it'll use the modified non-Bikeshed sources to republish the RD.

In practice this seems unlikely to happen since we'd only modify RDs within a short time window and ideally never, but it's worth keeping in mind and maybe researching a bit what a more robust system would require.

[annevk]

Let's dupe this into whatwg/meta#161 for now.