cmake: directly define project version

[lucascolley]

See explanation and discussion at https://lists.boost.org/Archives/boost/2024/09/257793.php. cc @pdimov, x-ref scipy/scipy#21270

[mborland]

These changes look fine to me. The real question is will this be better for you? About 2 years ago when Scipy moved to using the Boost.Math standalone submodule (scipy/scipy#17432) you were able to start tracking the head commit of math instead of having to wait for the tri-annual release. Won't this put you back in a position where you are waiting potentially months for fixes or features? Example. I am pretty sure your implementation of the Landau distribution relies on ours right now (through _ufuncs), but the Landau distribution has not made any official release for us yet and will not until 11 December at the earliest.

[lucascolley]

The real question is will this be better for you? About 2 years ago when Scipy moved to using the Boost.Math standalone submodule (scipy/scipy#17432) you were able to start tracking the head commit of math instead of having to wait for the tri-annual release. Won't this put you back in a position where you are waiting potentially months for fixes or features?

Indeed, there is a trade-off to be made there. We haven't made that decision yet, but these changes will enable us to go either way.

[mborland]

Indeed, there is a trade-off to be made there. We haven't made that decision yet, but these changes will enable us to go either way.

Sounds good. Since this adds a maintenance burden of like 30 seconds every 4 months I am not opposed.

[lucascolley]

Thanks!