Revert "Revert "cl refactor: add preload stage""

[xushiwei]

Reverts #895

[codecov]

Codecov Report

❌ Patch coverage is 86.89320% with 27 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.54%. Comparing base (7dfb6ab) to head (e9000d0).

Files with missing lines	Patch %	Lines
cl/preload.go	87.83%	18 Missing ⚠️
cl/class.go	70.37%	8 Missing ⚠️
cl/type.go	83.33%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #901      +/-   ##
==========================================
- Coverage   89.38%   88.54%   -0.85%     
==========================================
  Files          20       21       +1     
  Lines        1922     1946      +24     
==========================================
+ Hits         1718     1723       +5     
- Misses        204      223      +19     

Flag	Coverage Δ
llgo-tests	88.54% <86.89%> (-0.85%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 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.

[fennoai]

Review: re-apply class/function loading refactor

This PR re-applies the two-phase load refactor in cl/ (unifying funcObj/templateClass into overloadObj, splitting preload/load/compile phases via the new cl/preload.go). The build passes (go build ./cl/...) and the refactor logic reads soundly: newClassCtx nil handling is correctly guarded in loadClass/loadTemplateClass, the two-queue compile() flow runs each unit once, and preloadClass's NumTemplateArguments() >= 0 check is correct (-1 = not a template).

Main findings are inline. In short:

• addOveerloadObj is misspelled (double "e") across its definition, 4 call sites, and the debug log string in a brand-new file — worth fixing now before it gets entrenched.
• Two LLVM tests were silently disabled by renaming to _Test..., one with a leftover debug log.Println. No comment or tracking reason; t.Skip("reason") would keep the skip visible.
• A bare panic on a nested partial specialization (cl/class.go:352) replaces the previous graceful handling — a valid-but-unsupported C++ construct now hard-crashes the generator with no source location.

Non-blocking observations:

• Performance: newClassCtx adds a second clang.VisitChildren pass over each class's members (the first being initClassTypeEx, where loadClassMember is now a no-op for methods). This appears intentional — method USRs must be preloaded before out-of-line definitions resolve in loadOutsideMethod — but it does double the per-class child walk; worth confirming it's needed for classes with no out-of-line methods.
• Docs: cl/class.go removed the "class name acts like a namespace" rationale comments on the static-VarDecl and nested-class branches, which leaves the "same reason as a named nested class/struct above" cross-reference on the nested-union branch dangling. Also, compileFuncOrMethod's doc still describes a cls parameter that is now named this.

[fennoai]

Renaming to _TestLLVM_DenseMapInfo removes this from the Go test runner with no compile/lint signal and no explanation — coverage is lost invisibly. If intentional, add a comment/tracking issue, or use t.Skip("reason") so the skip is visible in test output.