Fix CI

[majosm]

Questions for the review:

• Is the scope and purpose of the PR clear?
  • The PR should have a description.
  • The PR should have a guide if needed (e.g., an ordering).
• Is every top-level method and class documented? Are things that should be documented actually so?
• Is the interface understandable? (I.e. can someone figure out what stuff does?) Is it well-defined?
• Does the implementation do what the docstring claims?
• Is everything that is implemented covered by tests?
• Do you see any immediate risks or performance disadvantages with the design? Example: what do interface normals attach to?

[matthiasdiener]

Instead of upgrading the version for everyone, could we just upgrade it for CI?

[majosm]

@matthiasdiener This is the Mac issue I was talking about earlier: #1103.

[majosm]

Seems to be working now. The doublemach mesh stuff resulted from inducer/meshmode#477; the two doublemach examples were previously generating slightly different meshes and "writing" them to the same file (but before that PR the output file name parameter didn't actually do anything, so it was never an issue).

[majosm]

@matthiasdiener Disabling SVM on non-NVIDIA seems OK as far as I can tell. Timestep times:

	Mac (KS2D)	Lassen CPU (KS3D)
pocl5	-	1.51 s
pocl6+SVM	0.548 s	1.91 s
pocl6+CL buffer	0.092 s	1.42 s

[majosm]

One of the times that I ran CI here I got a failure in the thermally-coupled-tpe examples (output here). I haven't been able to reproduce it since, so I'm going to let it go for now.

[matthiasdiener]

@matthiasdiener Disabling SVM on non-NVIDIA seems OK as far as I can tell. Timestep times:

Mac (KS2D) Lassen CPU (KS3D)
pocl5 - 1.51 s
pocl6+SVM 0.548 s 1.91 s
pocl6+CL buffer 0.092 s 1.42 s

Casual 6x speedup with CL buffers 🤷‍♂️