Implement example geometries for all detectors

[JamesWrigley]

We discussed this ages ago on Zulip (see the Default geometries? topic) but never got around to implementing it. I think most of the changes here are fairly straightforward, except for the LPD mini and JF which can have a variable number of modules. Here's what they look like:

The odd thing compared to the examples is that the origin is not centered. I tried centering it but got very confused and gave up :shrug: Geometry is quite bamboozling...

Some other things:

• You could argue that PNCCDGeometry and the epix classes don't need a .example() because some of their existing methods already have reasonable defaults. I'd say that it's good to have .example() methods anyway for consistency, but I don't feel very strongly about it.
• Going a bit further than .example(), the epix classes now have 'real' geometry methods .monolithic_geometry() and .pair_geometry() for convenience.

[JamesWrigley]

Partially done in 370ec3b, new geometry:

I didn't do it for detectors like the PNCCD and epix's for aesthetics :stuck_out_tongue:

[takluyver]

Thanks, I think this is a good idea. I agree with the change you've already made to put the origin at a corner for some of the less regular detectors.

For LPD mini, let's just put the N modules in a single column, rather than two columns. I think they've only got about 3 modules for this, and one of those is a different type of sensor, so you can't exactly use them all together as 'one detector'.

[JamesWrigley]

Gotcha, fixed in b1190eb:

[JamesWrigley]

Hmm, we seem to have run into this in CI: https://github.com/silx-kit/pyFAI/issues/1947 FWIW, all the tests pass locally for me.

[takluyver]

It sounds like we can probably work around it by pinning numexpr in the constraints file.

[takluyver]

Thanks, LGTM (once CI is sorted)