feat: Report processor switches to osqp when highs solver fails to converge.

[ple13]

No description provided.

[wfa-reviewable]

This change is 

[kungfucraig]

You should have some tests.

Reviewed 2 of 3 files at r1, all commit messages.
Reviewable status: 2 of 3 files reviewed, 3 unresolved discussions (waiting on @ple13)

---

src/main/python/wfa/measurement/reporting/postprocessing/noiseninja/solver.py line 155 at r2 (raw file):

  def _solve_with_initial_value(self, solver_name, x0) -> Solution:
    problem = self._problem()
    solution = solve_problem(problem, solver=solver_name, initvals=x0, verbose=False)

nit: Maybe just write out "initial_values"

Otherwise at least put an underscore between the words.

---

src/main/python/wfa/measurement/reporting/postprocessing/noiseninja/solver.py line 158 at r2 (raw file):

    return solution

  def _problem(self):

While you're in here you may as well add some logging about each attempt.

---

src/main/python/wfa/measurement/reporting/postprocessing/noiseninja/solver.py line 189 at r2 (raw file):

          attempt_count += 1

      # If the highs solver does not converge, switch to the osqp solver which

So solve with HIGHS for 10 attempts and if those fail, try just once with OSPQ?

[ple13]

I've added a unit test to verify that osqp will be executed when highs does not converge.

Reviewable status: 2 of 4 files reviewed, 2 unresolved discussions (waiting on @kungfucraig)

---

src/main/python/wfa/measurement/reporting/postprocessing/noiseninja/solver.py line 155 at r2 (raw file):

Previously, kungfucraig (Craig Wright) wrote…

nit: Maybe just write out "initial_values"

Otherwise at least put an underscore between the words.

Done.

---

src/main/python/wfa/measurement/reporting/postprocessing/noiseninja/solver.py line 158 at r2 (raw file):

Previously, kungfucraig (Craig Wright) wrote…

While you're in here you may as well add some logging about each attempt.

I have the PR#1878 to handle the log. I'll have the logs included in that PR (after this PR has been merged). WDYT?

---

src/main/python/wfa/measurement/reporting/postprocessing/noiseninja/solver.py line 189 at r2 (raw file):

Previously, kungfucraig (Craig Wright) wrote…

So solve with HIGHS for 10 attempts and if those fail, try just once with OSPQ?

Done.

[kungfucraig]

Reviewable status: 2 of 4 files reviewed, 2 unresolved discussions (waiting on @ple13)

---

src/main/python/wfa/measurement/reporting/postprocessing/noiseninja/solver.py line 189 at r2 (raw file):

Previously, ple13 (Phi) wrote…

Done.

Sorry. My comment was ambiguous.

Why try 10 times with the other one and just once with this one?

With respect to the comment you added, you noted it was more robust, so you should explain why we don't use it to start with.

[ple13]

Reviewable status: 2 of 4 files reviewed, 2 unresolved discussions (waiting on @kungfucraig)

---

src/main/python/wfa/measurement/reporting/postprocessing/noiseninja/solver.py line 189 at r2 (raw file):

Previously, kungfucraig (Craig Wright) wrote…

Sorry. My comment was ambiguous.

Why try 10 times with the other one and just once with this one?

With respect to the comment you added, you noted it was more robust, so you should explain why we don't use it to start with.

I'm sorry I didn't explain this better. The reason why I didn't try 10 times is because I thought the osqp solver was deterministic given the initial values. However, it is also a randomization process due to some non-deterministic internal states. I've added more comments to explain why we go for the highs solver first even though osqp is more robust (page 3 in https://arxiv.org/pdf/1711.08013 discusses first-order method (like osqp) vs others and suggests that osqp is less accurate.)