Avoid undefined-reference errors for BigFloat/BigInt

[kshyatt]

If you create a Vector{BigFloat}(undef, N), because BigFloat is "secretly" itself a vector, once you fill it you can hit undefined reference errors with the naive fill!. This adds a special path for these big numbers and some tests.

[lkdvos]

This is actually a really annoying thing: in Julia 1.12 there is an easy fix by just using map!(Returns(val), A) (single-arg), but that is 1.12+ only. I would slightly prefer if we could make that work anyways instead of having to special case here, but it might require being a bit careful with the 1.10 versions

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/mapreduce.jl	93.38% <100.00%> (+0.56%)	⬆️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[kshyatt]

I would slightly prefer if we could make that work anyways instead of having to special case here, but it might require being a bit careful with the 1.10 versions

Do we feel strongly enough about this to do it in this PR, or leave a comment and follow up?

[lkdvos]

I tried doing what I had in mind here, let me know if this is what you had in mind?

Slightly importantly this is actually not entirely following the Base implementation, which is actually calling map!(f, A) = map!(f, A, A), which has some slightly weird effects about number of arguments that end up being supplied to the function. (i.e. map!(sin, A) would now fail because sin() doesn't work). To be completely honest, I think this is a better design anyways, but it might be too opinionated?

[edit] while writing this I actually changed my mind, this last example seems silly to me, so probably better to follow base anyways, my bad.