Lazily init cholmod

[IanButterworth]

On top of #625

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 83.95%. Comparing base (7b6e810) to head (fac1548).
Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #626      +/-   ##
==========================================
- Coverage   83.97%   83.95%   -0.03%     
==========================================
  Files          12       12              
  Lines        9265     9265              
==========================================
- Hits         7780     7778       -2     
- Misses       1485     1487       +2     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[rayegun]

Hmm I wonder if we should build SuiteSparse against libjulia for the jl_<GC> functions in Ygg. This will now no longer use jl_malloc for a number of JLLs including Sundials as well as any of the other SuiteSparse packages

[ViralBShah]

That would be a good idea!

[staticfloat]

Why did this change anything? The pointers to those functions should still be getting set in that once per process block, no?

[rayegun]

My understanding is that __init__ would be called by Julia always at load time, therefore cholmod will be ready for any downstream JLLs. Whereas now you must call getcommon which may not occur until you actually use cholmod. This is really something we should stick in a prologue on the JLL I think?

[IanButterworth]

I think additionally, I did wonder if this change is safe for deserialized cholmod objects?