FE-1341: Give experiment compute backends one interface

[kube]

🌟 What is the purpose of this PR?

Adds @hashintel/petrinaut-core/experiments: one interface both compute backends satisfy, and a registry that picks between them.

ExperimentsProvider currently hardcodes the choice. #9179 registers a backend against this interface instead of rewiring the provider.

Refactor only. No behaviour change, no changeset.

🔗 Related links

• FE-1341 (internal)
• FE-1340 (internal), the second backend

🚫 Blocked by

• FE-1237: Add optional per-place token capacity #9177

🔍 What does this change?

Both paths already produce a MonteCarloExperiment, and consumers drive one with no branching. That half is untouched. This adds four types:

Type	What it is
ExperimentRequest	What to compute, as serializable data. No worker factory, no ODE method, no signal, no callbacks: those describe how, and belong to backend construction or per-call options
ExperimentAssessment	A backend's answer to "can you run this?", as a value. Blockers carry code, optional itemId, and an origin
ExperimentBackend	assess settles the net without acquiring a device or worker pool, then returns an instantiate() closure carrying that work forward
selectExperimentBackend	Walks registrations in preference order, records each refusal, treats a failed instantiation as a refusal

Blocker origins drive what a UI does:

Origin	Action
model	Edit the net
configuration	Edit the experiment
environment	Hide the option
capacity	Retry, or use fewer runs

flowchart TD
  R[registrations] --> L[load] --> A{assess}
  A -->|blockers| N[record refusal] --> R
  A -->|eligible| I{instantiate}
  I -->|environment / capacity| N
  I -->|handle| D[run]

Loading

load is deferred, so a heavy backend stays out of a bundle that never uses it.

Adds content/experiments/backend-selection.mdx and the core.experiments layer.

Pre-Merge Checklist 🚀

🚢 Has this modified a publishable library?

This PR:

• does not modify any publishable blocks or libraries, or modifications do not need publishing

📜 Does this require a change to the docs?

The changes in this PR:

• require changes to docs which are made as part of this PR

🕸️ Does this require a change to the Turbo Graph?

The changes in this PR:

• do not affect the execution graph

⚠️ Known issues

One backend is registered here, so assess never refuses and the blocker types are unexercised until #9179.

Worker pools and GPUDevice are still per experiment. Reuse depends on the worker resetting frame and metric state; if it does not, results stay plausible rather than visibly wrong.

🐾 Next steps

Supply backends through a React context, so an embedder can register its own.

🛡 What tests cover this?

select-experiment-backend.test.ts: 9 cases covering preference order, deferred loading, refusal recording, instantiation failure, blocker ordering, unavailable backends, module load failure, and per-requirement request building.

❓ How to test this?

Run an experiment. Behaviour is unchanged.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
hash	Ready	Preview	Aug 15, 2026 1:13am
petrinaut	Error		Aug 15, 2026 1:13am

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
hashdotdesign-tokens	Ignored	Preview	Aug 15, 2026 1:13am