HSGP misc fixes

[bwengals]

Description

• Fix centering calculation (replace - with +) (nice catch @ulfaslak!)
• Use x_range instead of x in approx_hsgp_hyperparams. I don't think the full x is needed anymore, could be missing something though.
• Add more detail in docstring of approx_hsgp_hyperparams about usage.
• Renamed Xs to X. The "s" stood for "scaled", for when the user had to center or mean subtract the incoming X. Now this is done automatically, so it can just be called X. Though minor, it is a breaking change. Would it be good to add a deprecation warning and keep the Xs syntax for a while?

Also, didn't make this change, but what do people think about not returning S from approx_hsgp_hyperparams, so it would just return m and c? The idea was that returning S here would be helpful for the user when centering things, but that's not really necessary anymore.

---

📚 Documentation preview 📚: https://pymc--7342.org.readthedocs.build/en/7342/

[codecov]

Codecov Report

Attention: Patch coverage is 70.58824% with 5 lines in your changes are missing coverage. Please review.

Project coverage is 92.37%. Comparing base (508a134) to head (7640a18).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7342      +/-   ##
==========================================
- Coverage   92.37%   92.37%   -0.01%     
==========================================
  Files         102      102              
  Lines       17208    17206       -2     
==========================================
- Hits        15896    15894       -2     
  Misses       1312     1312              

Files	Coverage Δ
pymc/gp/hsgp_approx.py	88.39% <70.58%> (-0.13%)	⬇️

[AlexAndorra]

Thannks @bwengals !

Renamed Xs to X. [...] Would it be good to add a deprecation warning and keep the Xs syntax for a while?

I think it's fine because it's gonna get used more now that we have examples and tutorials up, so better do the change now than later.

What do people think about not returning S from approx_hsgp_hyperparams, so it would just return m and c?

Agreed, we can get rid of that, as it makes the API simpler. Then we can probably get rid of the NamedTuple?
Note that all those changes need to be added to the tutorial part 1 that we just published.

[AlexAndorra]

Almost all good @bwengals !
Just one change I highlighted, plus getting rid of returning S in the approx function (as well as updating the NB on pymc-examples)

[juanitorduz]

Small suggestion: Could a unit test be added to check this so that we are sure future changes won't break the signs? We could just wrap this little logic into a small axillary function so that we can easily test it.

[juanitorduz]

This looks great! Thanks! I agree with Alex's comments. I just left a small suggestion above :)

[AlexAndorra]

Just have a last question @bwengals

[bwengals]

Sounds good, I can make these changes on Friday

[AlexAndorra]

I got some time, so I just pushed the change to the formula for S @bwengals 🍾
Feel free to merge if it looks good and tests pass