Allow passing kwargs on to Libtask (when it's an AbstractTuringLibtaskModel...)

[penelopeysm]

This PR modifies LibtaskExt to allow sampling from Turing models with keyword arguments with SMC or PG. See TuringLang/Turing.jl#2660 for usage.

It's obviously quite ugly, the problem is that even though Turing is the only library that subtypes the abstract type, the abstract type has to be defined in AdvancedPS so that the LibtaskExt can dispatch on it. Ideally, that abstract type + any relevant methods inside LibtaskExt should just be shifted into Turing proper. That's for another time.

[yebai]

Yes, we'd like to merge the AdvancedPS code to Turing.jl. Libtask was holding us back; now, it is ready.

[github-actions]

AdvancedPS.jl documentation for PR #118 is available at:
https://TuringLang.github.io/AdvancedPS.jl/previews/PR118/

[penelopeysm]

Hi @yebai, sorry just tidying this up before I go away :)

I am happy to try to work on upstreaming the necessary parts to Turing next year, but that might take more time, and this PR will unblock keyword arguments on SMC/PG on Turing (TuringLang/Turing.jl#2660). So I think it makes sense to merge this first. I tested locally and it all works as intended, so happy for it to be reviewed now (cc also @mhauru).

[codecov]

Codecov Report

❌ Patch coverage is 42.85714% with 4 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (main@e100352). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
ext/AdvancedPSLibtaskExt.jl	42.85%	4 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #118   +/-   ##
=======================================
  Coverage        ?   95.41%           
=======================================
  Files           ?        8           
  Lines           ?      436           
  Branches        ?        0           
=======================================
  Hits            ?      416           
  Misses          ?       20           
  Partials        ?        0           

☔ 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.