Add iteration-count accessors to SolvedModel

[GermanHeim]

After a solve, HiGHS exposes per-algorithm iteration counts through its solution-info values (simplex_iteration_count, ipm_iteration_count, etc.).
Right now, there's no safe way to read them: a caller has to reach for SolvedModel::as_mut_ptr() and call the raw Highs_getIntInfoValue FFI itself.

With this PR, we add five new public methods on SolvedModel, mirroring the existing accessors:
simplex_iteration_count(), ipm_iteration_count(), qp_iteration_count(), pdlp_iteration_count(), and crossover_iteration_count().

I also addeda small private int_info_value(name) helper that wraps Highs_getIntInfoValue + try_handle_status.

[lovasoa]

can this fail in normal operations ? if yes, then return a Result. Otherwise, use expect instead of unwrap, and explain why this can't fail clearly

[GermanHeim]

Highs_getIntInfoValue returns kError in only two cases, an unknown info name or a name whose info isn't HighsInt-typed. Both depend only on the name argument. Every caller passes a valid HighsInt key (c"simplex_iteration_count", etc.), so the only way to trigger it is a code change that introduces a bad literal.

I based this accessor based on others in the library, for example mip_gap, primal_solution_status. Should we also switch to an expect there or is it fine as it is?

[lovasoa]

If the safety boundary is in the callers, let's put the .expect(...) there. int_info_value can be public, return a Result, and document when it returns an error. The accessor functions can be trusted and not return a Result because the parameter they pass is trusted

[GermanHeim]

Done, and extended the same to the existing mip_gap()/primal_solution_status().

[lovasoa]

excellent, great contribution, thanks @GermanHeim !