Type-stabilize broadcast over tuples and scalars

[pabloferz]

No description provided.

[pabloferz]

The only problem with this would be the codegen for large tuples, but that should be fixed on the ntuple end anyway.

[Sacha0]

@pabloferz, this change looks great! Addressing the potential code blowup issue in ntuple sounds like a good plan if possible; perhaps something like the Any16 trick for map over tuples would serve?

@nanosoldier runbenchmarks(ALL, vs = ":master")

[Sacha0]

Ref. #20817 (comment), which this change might address by using ntuple(..., Val{N}) rather than ntuple(..., n).

[pabloferz]

Seems like @nanosoldier cannot be triggered from a review summary.

@nanosoldier runbenchmarks(ALL, vs = ":master")

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @jrevels

[Sacha0]

The sole possible regression nanosoldier identified is almost certainly noise. Cheers!