First step to target fully source gen implementation

[kant2002]

I extract simple model for Refit client representation withing source gen That to allow facilitate process

• Parse Syntax (done by Roslyn itslef)
• Build internal model
• Write based on model only, so logic would be as simple as possible

Also that allows write more robust test if we test just internal model extraction from symbols. Otherwise if we will capture generated output, if we decide improve generation it will produce a lot of churn. I do not agains churn like that, but not sure that's would be appropriate.

Related to #1389

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 83.66%. Comparing base (6ebeda5) to head (9bdb28b).
Report is 170 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1746      +/-   ##
==========================================
- Coverage   87.73%   83.66%   -4.08%     
==========================================
  Files          33       36       +3     
  Lines        2348     2442      +94     
  Branches      294      345      +51     
==========================================
- Hits         2060     2043      -17     
- Misses        208      316     +108     
- Partials       80       83       +3     

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

[glennawatson]

Thanks @kant2002, and the direction here is sound (a model-based generation step for simpler, more testable codegen). It's been conflicting and stalled for ~18 months and the source-gen/AOT work has moved on since, so closing to keep things tidy rather than leave it lingering. The idea is still welcome — please reopen or rebase a fresh PR against current main if you'd like to pick it up.