The general intent here is to support basic use of `i32` and similar
integer types in min_prelude tests, without all the various arithmetic
support.

This is in support of #5547 and #5549. However, I was originally asking
for this to be reviewed as part of #5546 rather than either of those
PRs, and I've retracted #5546 due to [the discussion on Discord about
test change
complexity](https://discord.com/channels/655572317891461132/655578254970716160/1377406366200758293).
So, to try to still get this in (and then merge the already-approved
#5547 and #5549), splitting out this file.

Note this is actually the form of the prelude in #5549, which was adding
negate -- it felt better to me to add that together as long as I'm
splitting it out.