Prep for v1.40.0

[odow]

TODO

• Fix SetMapBridge with complex type #2733
• [Nonlinear.ReverseAD] Fix use of reinterpret instead _reinterpret_unsafe #2744
• Add docstring to _reinterpret_unsafe #2745
• Add import MathOptInterface as MOI to DocTestSetup #2746

Basic

• version field of Project.toml has been updated
  - If a breaking change, increment the MAJOR field and reset others to 0
  - If adding new features, increment the MINOR field and reset PATCH to 0
  - If adding bug fixes or documentation changes, increment the PATCH field

Documentation

• Add a new entry to docs/src/changelog.md, following existing style

Tests

• https://github.com/jump-dev/MathOptInterface.jl/actions/runs/14745183582
• https://github.com/jump-dev/MathOptInterface.jl/actions/runs/14784860409
• If new tests were added, ensure that MOI.Test.version_added is implemented.

[odow]

@blegat a bunch of solver-test failures from your ReverseAD changes. I'll take a look this afternoon.

[blegat]

Clarabel is failing its native optimizer tests and it was already failing at the last release so it's fine (it's also been reported oxfordcontrol/Clarabel.jl#198)
For BilevelJuMP, it was working at the last release so did we break it ? #2719

[odow]

I think BilevelJuMP is just a flakey test because @joaquimg sets a time limit of 5 seconds. I guess a big portion of that is precompilation.

[joaquimg]

I will look into it.

[odow]

Ah. It'll be because I'm not precompiling a solve that uses ReverseAD in Ipopt: https://github.com/jump-dev/Ipopt.jl/blob/0acb50a719db990fc5529743d44e7b5fa59c1ef6/ext/IpoptMathOptInterfaceExt/IpoptMathOptInterfaceExt.jl#L21-L77