feat: add pygfx-backed histogram

[gselzer]

This is a repeat of #76, ready for eyes again.

There are a few problems I'm aware of, which I'm tempted just solve in separate PRs:

• There's "shimmering" when you pan/zoom the histogram canvas, due to aliasing. PyGFX's antialias solution of renderer.pixel_ratio was not sufficient to fix this on my machine.
  • Note that the VisPy histogram also has this issue, but it is only seen when zooming.
• When you change the data, the pan/zoom resets. This is also a problem for the VisPy histogram.

[codecov]

Codecov Report

Attention: Patch coverage is 56.15942% with 121 lines in your changes missing coverage. Please review.

Project coverage is 75.69%. Comparing base (133f7c1) to head (f39c5fc).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/ndv/views/_pygfx/_histogram.py	55.83%	121 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #105      +/-   ##
==========================================
+ Coverage   72.43%   75.69%   +3.26%     
==========================================
  Files          48       49       +1     
  Lines        4622     4897     +275     
==========================================
+ Hits         3348     3707     +359     
+ Misses       1274     1190      -84     

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

[tlambert03]

completely works for me!

nit re fps

[tlambert03]

maybe remove show_fps or guard it behind an env var or something?

[gselzer]

Oh, good catch - will remove, I'd rather more thoroughly address that across the repo in #47

[tlambert03]

added a basic smoke test

[tlambert03]

one test still consistently failing with pygfx and pyside on windows: https://github.com/pyapp-kit/ndv/actions/runs/12856369561/job/35842981229?pr=105

not 100% sure it's the histogram test, but it seems like it's happening right about the same time

[tlambert03]

was able to reproduce the test segfault on a real windows 3.12, pyside, pygfx about 50+% of the time. So I do think it's a real bug. But since we're already only partially supporting pyside, let's go ahead and skip it

[tlambert03]

i'm going to merge this, but I think the additional test I added here, though just a simple smoke test, is causing more random test failures. It could be two things: either something related to the histogram test itself, or simply related to the fact that the new test is actually the first test added that (re-)uses the backend application for a second time. That does come with a whole host of possible gotchas. But, since that's unrelated to this PR, and since we were likely to hit it eventually anyway, I'm going to merge this, and we may have to deal with a brief period of test flakiness while I figure that out.

thanks for your work here!