Bypass cyclic emulator state indexing

[DennisWayo]

Summary

This is a small follow-up to the emulator/Qibocal baseline enabled by qiboteam/qibocal #1551.

It updates cyclic emulator result extraction to marginalize probabilities directly over the measured Hilbert-space component instead of first building the full computational-state index table and masking it.

The behavior is unchanged at the result interface:

• non-ground states are still classified as population 1;
• INTEGRATION results still return the expected (I, Q) shape;
• sweep axes are preserved.

Context

This addresses the marginalization direction described in qiboteam/qibolab#1216: for non-shot/cyclic result processing, there is no need to go through sampled-state machinery or full-state index conversion when probabilities can be reshaped to the system Hilbert-space dimensions and marginalized directly.

I kept this separate from the local #1413 work because #1413 depends on the confusion-matrix path introduced in draft qiboteam/qibolab #1401, which has not landed yet.

Validation

• pytest tests/instruments/emulator/test_results.py -q
• pytest tests/instruments/emulator/test_emulator.py tests/instruments/emulator/test_results.py -q
• pytest tests/instruments/emulator -q
• ruff check src/qibolab/_core/instruments/emulator/results.py tests/instruments/emulator/test_results.py
• ruff format --check src/qibolab/_core/instruments/emulator/results.py tests/instruments/emulator/test_results.py
• Qibocal Rabi smoke on merged Qibocal main platform: qubit acquisition + fit passed
• Qibocal Rabi smoke on merged Qibocal main platform: qutrit acquisition + fit passed

[lballerio]

Hi Dennis, thanks for your contribution!
For me is almost fine, I would ask you do the for loop inside the _marginalize_probability, this would be better for modularity and readability.
Also, for even more readability, maybe add some more comments especially in the more convoluted lines.
I'll move this PR to draft, but I think apart for these small suggestions I think it will be soon good to merge for me.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 39.41%. Comparing base (6530bdf) to head (932ae74).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1505      +/-   ##
==========================================
+ Coverage   39.34%   39.41%   +0.06%     
==========================================
  Files         115      115              
  Lines        6145     6152       +7     
==========================================
+ Hits         2418     2425       +7     
  Misses       3727     3727              

Flag	Coverage Δ
unittests	39.41% <100.00%> (+0.06%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[lballerio]

I guess here you are summing all states >=1...I think you can also move this part in the function

[DennisWayo]

Thanks @lballerio, this makes sense. I moved the per-measurement loop into _marginalize_probability(), and the helper now also makes the non-ground-state classification explicit there. I also added the comment around states outside the ground state being classified as 1, so the intent is clearer at the point where the summation happens.

[DennisWayo]

Thanks @lballerio , I added the comment at that point.

It now states that the non-ground-state sum is intentionally done before stacking because measured subsystems may have different dimensions.

[lballerio]

Hi @DennisWayo, I tested your code for a bit and everything seems working, so for me is approved and able to merge.