Implement AD testing and benchmarking (hand rolled)

[penelopeysm]

One of two options. The other one at #883.

This PR implements functionality for testing and benchmarking AD. It is largely copied over from my ModelTests repo where I've been playing around with this.

Closes #869

What does it contain?

It basically adds one function DynamicPPL.TestUtils.AD.run_ad. See the docstring for more info.

Why not an extension?

The only new dependencies are Statistics, which is stdlib, and Chairmarks, which itself has no non-stdlib dependencies. I therefore consider it unnecessary to add an extension (which would bring a number of drawbacks, e.g. reduced discoverability as users have to load the trigger packages themselves, us having to faff around with functions declared in src/ and extended in ext/, ...)

Why do I like this one more?

See #883.

[github-actions]

Benchmark Report for Commit b107b92

Computer Information

Julia Version 1.11.4
Commit 8561cc3d68d (2025-03-10 11:36 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 4 × AMD EPYC 7763 64-Core Processor
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, znver3)
Threads: 1 default, 0 interactive, 1 GC (on 4 virtual cores)

Benchmark Results

|                 Model | Dimension |  AD Backend |      VarInfo Type | Linked | Eval Time / Ref Time | AD Time / Eval Time |
|-----------------------|-----------|-------------|-------------------|--------|----------------------|---------------------|
| Simple assume observe |         1 | forwarddiff |             typed |  false |                  9.3 |                 1.6 |
|           Smorgasbord |       201 | forwarddiff |             typed |  false |                601.0 |                43.1 |
|           Smorgasbord |       201 | forwarddiff | simple_namedtuple |   true |                423.7 |                46.3 |
|           Smorgasbord |       201 | forwarddiff |           untyped |   true |               1180.2 |                29.1 |
|           Smorgasbord |       201 | forwarddiff |       simple_dict |   true |               3827.5 |                20.5 |
|           Smorgasbord |       201 | reversediff |             typed |   true |               1442.6 |                29.1 |
|           Smorgasbord |       201 |    mooncake |             typed |   true |                919.0 |                 5.3 |
|    Loop univariate 1k |      1000 |    mooncake |             typed |   true |               5402.5 |                 4.1 |
|       Multivariate 1k |      1000 |    mooncake |             typed |   true |               1080.6 |                 8.3 |
|   Loop univariate 10k |     10000 |    mooncake |             typed |   true |              59302.8 |                 3.7 |
|      Multivariate 10k |     10000 |    mooncake |             typed |   true |               8930.4 |                 9.6 |
|               Dynamic |        10 |    mooncake |             typed |   true |                133.8 |                11.8 |
|              Submodel |         1 |    mooncake |             typed |   true |                 25.0 |                 8.0 |
|                   LDA |        12 | reversediff |             typed |   true |                382.9 |                 6.1 |

[codecov]

Codecov Report

Attention: Patch coverage is 68.96552% with 9 lines in your changes missing coverage. Please review.

Project coverage is 84.80%. Comparing base (c7bdc3f) to head (ef5a1ce).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/test_utils/ad.jl	68.96%	9 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #882      +/-   ##
==========================================
- Coverage   84.92%   84.80%   -0.13%     
==========================================
  Files          34       35       +1     
  Lines        3814     3843      +29     
==========================================
+ Hits         3239     3259      +20     
- Misses        575      584       +9     

☔ 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.

[coveralls]

Pull Request Test Coverage Report for Build 14448023153

Details

• 0 of 28 (0.0%) changed or added relevant lines in 1 file are covered.
• 3 unchanged lines in 1 file lost coverage.
• Overall coverage decreased (-0.1%) to 84.892%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/test_utils/ad.jl	0	28	0.0%

Files with Coverage Reduction	New Missed Lines	%
src/varinfo.jl	3	83.83%

Totals
Change from base Build 14392526425:	-0.1%
Covered Lines:	3259
Relevant Lines:	3839

---

💛 - Coveralls

[yebai]

Thanks, @penelopeysm. This looks good!