Support SC/SI variables

[GermanHeim]

This PR adds support for semicontinuous and semi-integer variables.
We add an Integrality enum (Continuous, Integer, SemiContinuous, SemiInteger) and methods to create semicontinuous/semi-integer vars, which take value 0 or fall within their bounds.

[GermanHeim]

Hi @lovasoa, should I also bump highs-sys to v1.15.0 in this PR?

[lovasoa]

Yes, I think this warrants a version bump. Sorry for the delay.

For the pr itself: do you think a separate knob to add zero to the solution domain would be clearer ?

Something like

struct SolutionDomain {
   /// Allow only integers
   is_integral: boolean,
  
  /// Allow the variable to take the value 0 even if it's not in its [lower, upper] bounds
  allow_zero: boolean
}

[jajhall]

HiGHS v1.15.1 is current

[GermanHeim]

I think allow_zero is redundant. We normally think of these as SC/SI vars, which the enum already represents directly. I'd probably keep the current API unless there's a strong preference for exposing an explicit allow_zero flag.

[lovasoa]

@GermanHeim ok

[GermanHeim]

@lovasoa Just as a heads up, we are still using HIGHS v1.14.
Should we update to v1.15.1?

[lovasoa]

we are declaring a minimal supported version of highs-sys of

highs-sys = "1.14.3"

unless I'm mistaken, this is already compatible with v1.15.1.

highs is just a library, its job is to declare the version range of its dependencies it's compatible with. It does not decide what precise version of highs-sys will be installed when the dependency resolution algorithm runs.

Your job is to maintain the cargo.toml up to date so that it does not lie about which versions of highs-sys we support.

[GermanHeim]

Ah, understood.
I mistakenly thought highs-sys determined the version because we ship precompiled binaries, but that makes sense. Thanks for clarifying!

[GermanHeim]

Build highs*
Not precompiled

I got confused with russcip