we have a selection in our one-change (C0) history


 [S] C0

transaction comes in: -C0 R0 C0

So we need to map that selection, which sits _before_ C0, forward
through this new history (C0, -C0, R0). But it's only mapping through
two maps, which appear to be -C0 and R0.

Just mapping through C0 would also work, since mapping C0,-C0 is
guaranteed to not do a thing. Tests still pass. Is this generally
solid? There are guaranteed to be inverted version of the first
`rebased` steps at the start of the transaction. So when you looking
at item N, and mapping from the corresponding inverted step index, I
guess the thing at that index is always -N.step. Let's go with this.