Use OnlyAccsVarInfo in check_model

[penelopeysm]

Because the model is contextualised with InitContext, the call to check_model doesn't actually need anything more than an OnlyAccsVarInfo. This allows models that don't work well with Metadata* to still be checked for validity (note that the check, as defined in DynamicPPL, only uses DebugAccumulator, it does not actually use metadata in any form -- unless the model has DefaultContext as its leaf context -- which is not the case here).

(cc @PTWaade I think this is the reason why check_model failed)

* one might ask what models don't work well with metadata, but still otherwise work. There are a couple of things that come to mind: 1) parallel assume statements; and 2) special distribution types for which the from_[linked_]_vec_transform functions are not defined. These models should in principle work with Prior() sampling, which only ever uses OnlyAccsVarInfo. However, currently, they don't unless you disable check_model; this is the behaviour that this PR fixes.

[github-actions]

Turing.jl documentation for PR #2742 is available at:
https://TuringLang.github.io/Turing.jl/previews/PR2742/

[PTWaade]

This makes sense ! Lightning fast work Penelope - and I who thought you weren't working currently !

[penelopeysm]

ahhhh, it's only 3 lines so I made an exception 😄

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.51%. Comparing base (4dc7ad0) to head (f343bbb).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2742   +/-   ##
=======================================
  Coverage   86.51%   86.51%           
=======================================
  Files          20       20           
  Lines        1261     1261           
=======================================
  Hits         1091     1091           
  Misses        170      170           

☔ 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.