fix assumption that deep copied 0-field mutable structs compare equal

[KristofferC]

When running the tests for this package on the upcoming 1.3 release I got the error:

solverstring = "Alpine.UnsetSolver()"
Partitioning variable selection tests :: nlp3: Error During Test at /root/.julia/packages/Alpine/yThaY/test/solver.jl:18
  Got exception outside of a @test
  Unsupported MINLP local solver Alpine.UnsetSolver(); use a Alpine-supported MINLP local solver

I tracked it down to the following issue JuliaLang/julia#33359.

This PR works around the problem.

[codecov]

Codecov Report

Merging #132 into master will increase coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #132      +/-   ##
==========================================
+ Coverage   77.08%   77.09%   +<.01%     
==========================================
  Files          15       15              
  Lines        3155     3152       -3     
==========================================
- Hits         2432     2430       -2     
+ Misses        723      722       -1

Impacted Files	Coverage Δ
src/solver.jl	97.83% <ø> (+0.39%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4078aaf...6412ae0. Read the comment docs.

[KristofferC]

On more careful examination, this is the result of a bugfix in deepcopy with mutable structs being considered non-singleton (JuliaLang/julia#31825).

The lines at:

Alpine.jl/src/solver.jl

Lines 485 to 487 in 4078aaf

	deepcopy(nlp_solver),
	deepcopy(minlp_solver),
	deepcopy(mip_solver),

makes the check for == with empty_solver be false. I believe this is a correct bugfix for the problem shown by the test failure.

[harshangrjn]

Thanks @KristofferC for catching this.