Support arbitrary observation dimensions

[AlexanderFengler]

Stack

This is the observation-dimension child of #1266 and targets its rolling sink branch. It should be reviewed and merged into that sink, not directly into main.

Purpose

• derive generated observation width from the ordered physical response declaration
• reject callable simulator metadata that disagrees with that configured width
• support scalar and vector predictive outputs for dependency-free widths one through four

Changes

• capture one immutable configured _obs_dim before deadline handling can extend the technical response list
• thread that width through generated RV construction and predictive family coordinates
• omit the extra response coordinate for scalar choice-only outputs and derive vector coordinates dynamically
• replace fixed rt,response cleanup and DataFrame naming with one response-order-derived path
• preserve legacy model-name choice-only behavior, default two-column names, deadline width, RNG order, and cloudpickle reconstruction
• add synthetic width 1--4, mismatch, safe-mode chunking, sample_do, DataFrame, and serialization coverage

Complexity ledger

• production delta: 6 files, net +41 lines
• one stored integer and one shared private cleanup helper
• no new module, class hierarchy, model-name dispatch, or ssm-simulators metadata dependency
• no new public symbol; existing distribution helpers gain one optional expected-width argument
• duplicated fixed-width predictive cleanup branches are removed

Verification

• full non-slow suite: 991 passed, 10 subtests passed, 201 deselected
• selected slow posterior-predictive, prior-predictive, sample_do, and save/load canaries: 19 passed
• focused arbitrary-width and adjacent compatibility matrices
• Ruff check and format check on changed files
• mypy: 83 source files clean
• Pyrefly: 0 errors
• sdist and wheel build
• git diff --check

Refs #1265

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 7e18f123-4b6b-43ac-9183-5007b95923ec

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}