MR51: Allow Matplotlib to use system qhull >= 7.2.0

[sagetrac-gitlab-bot]

zlscherr (@zlscherr) opened a merge request at https://gitlab.com/sagemath/sage/-/merge_requests/51:

---

My system has qhull 8.0.0 installed.  Because Matplotlib
doesn't know about my system qhull it builds its own version.
This can cause runtime issues with Matplotlib trying to call
the wrong qhull library.  For example, this fixes the error that
```bash
$ ./sage -tp 8 --long src/sage/plot/plot3d/list_plot3d.py

produces

QH6249 qh_lib_check: Incorrect qhull library called.
Size of qhT for caller is 2896, but for library is 2792.

QH6256 qh_lib_check: Cannot continue.
Library 'qhull 7.2.0 (2015.2 2016/01/18)'
uses a static qhT (e.g., libqhull.so)

if the system has qhull 8.0.0 installed.

Depends on #31128

CC:  @slel

Component: **build**

Author: **Zachary Scherr**

Branch/Commit: **[`6a36667`](https://github.com/sagemath/sagetrac-mirror/commit/6a3666745f6c0ce268f142a8c6ae984ec4ff4dff)**

Reviewer: **Dima Pasechnik**

_Issue created by migration from https://trac.sagemath.org/ticket/31148_

[dimpase]

Reviewer: Dima Pasechnik

[dimpase]

comment:2

lgtm

[slel]

comment:3

Full name in Author field please.

[zlscherr]

Changed author from zlscherr to Zachary Scherr

[slel]

comment:5

Thanks!

[mkoeppe]

comment:6

The cffi upgrade that is on the ticket should be mentioned in the ticket description

[zlscherr]

comment:7

Oops, that's my fault since I'm still learning about contributing. That was leftover from the last ticket I submitted as I had merged it into my local develop branch. Is there an easy way to separate those or should I just submit a new ticket? Sorry about that.

[mkoeppe]

Dependencies: #31128

[mkoeppe]

comment:8

As the branch is on top of the one of #31128, it's OK to just add that ticket as a dependency. I have just done that.

(Best practice, however, would be to have independent branches for independent tickets.)

[zlscherr]

comment:9

Thanks.

Also, what is best practice for package dependencies? Matplotlib depends on numpy which depends on pkgconfig, which I use in this ticket. Does that mean I should explicitly add pkgconfig to Matplotlib's dependency file?

[mkoeppe]

comment:10

It would be clearer to add it explicitly;
but it is correct as is because we currently do not use any system Python packages - and hence the indirect dependencies are guaranteed to be present.

[dimpase]

comment:11

I just opened #31165 to deal with system-wide qhull.

We should also update our qhull.

[vbraun]

Changed branch from u/galois/mrs/51/matplotlib_qhull_fix to 6a36667