Fix performance of getparams on untyped vi

[penelopeysm]

Turing.Inference.getparams(model, vi) is really slow on untyped vi because it has to re-evaluate the model. It's much faster to convert to a typed vi before re-evaluating.

(In the medium-term future, once accumulators are in, we shouldn't need to rerun the model at all, but this at least makes prior sampling usable while we get that in.)

Closes #2604

[github-actions]

Turing.jl documentation for PR #2605 is available at:
https://TuringLang.github.io/Turing.jl/previews/PR2605/

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.59%. Comparing base (97c3bc7) to head (5d1f1e0).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2605      +/-   ##
==========================================
+ Coverage   85.57%   85.59%   +0.01%     
==========================================
  Files          22       22              
  Lines        1456     1458       +2     
==========================================
+ Hits         1246     1248       +2     
  Misses        210      210              

☔ 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.

[sunxd3]

fun!

[coveralls]

Pull Request Test Coverage Report for Build 15907755992

Details

• 2 of 2 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.02%) to 85.655%

---

Totals
Change from base Build 15901810849:	0.02%
Covered Lines:	1248
Relevant Lines:	1457

---

💛 - Coveralls