Zephyr layout coords fix

[mahdiehmalekian]

A bug in zephyr_layout in the file dwave-networkx/drawing/zephyr_layout.py is fixed so that if scale = scale, all positions fit within [0, scale] on the x-axis and [-scale, 0] on the y-axis.

A couple of tests are added to tests/test_zephyr_layout.py that would fail before fixing the bug. These test the range of x value and y-value produced by zephyr_layout for the default scale=1 and scale=10.

Similarly, for pegasus_layout and test_pegasus_layout.py.

Closes #239.

[codecov]

Codecov Report

Attention: Patch coverage is 6.45161% with 29 lines in your changes missing coverage. Please review.

Project coverage is 75.05%. Comparing base (deb095f) to head (e0ccc6f).
Report is 18 commits behind head on main.

Files with missing lines	Patch %	Lines
dwave_networkx/drawing/pegasus_layout.py	6.25%	15 Missing ⚠️
dwave_networkx/drawing/zephyr_layout.py	6.66%	14 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #236      +/-   ##
==========================================
- Coverage   75.96%   75.05%   -0.91%     
==========================================
  Files          31       31              
  Lines        2184     2213      +29     
==========================================
+ Hits         1659     1661       +2     
- Misses        525      552      +27     

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

---

🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[jackraymond]

@randomir - can you assign myself and @AndyZzzZzzZzz as reviewers?

[mahdiehmalekian]

@randomir Is this ready to be merged?

[jackraymond]

I understand it would be standard if numpy was imported at the top rather than within the layout function. Double check with @randomir

range != [0,1] in layout as described. The update corrects this problem, unit tests are good.

It is not clear to me why scale=1 must imply [0,1], when I use pos in combination with draw_networkx (for example) plots look equally good before and after this pull request (to my eye, superficially). Is there a specific example of when this causes unexpected behaviour? Happy to assume there is one and approve the request though.

[mahdiehmalekian]

You're correct that the plots don't look different before and after this PR. However the document says scale=1 gives values in [0,1], and this is a feature that was relied on in the layout algorithm (in dnx_layout in layout.py) and the bug in this code caused the layout algorithm to give poor results.

[mahdiehmalekian]

@randomir
Could you advise on numpy import?