refactor(workflows): decontaminate inference imports

[grzegorz-roboflow]

What does this PR do?

Remove 196 of 571 imports from inference/core/workflows in place, behind ports bound through the existing init_parameters path:

• AST import lint with a shrink-only baseline (571 -> 375)
• module-local stdlib loggers + structlog ProcessorFormatter adapter (parity test)
• ModelsProvider, WorkflowsCache, BackgroundTaskScheduler protocols; vendored LRUCache
• pure helpers vendored into workflows.utils with local exception/warning classes
• enterprise blocks loaded as a workflows_core-sourced plugin (prepended)
• Roboflow step error handlers moved to the server and passed explicitly at all four composition roots; workflows default is mapping-free legacy
• dead model_manager parameters removed from 30 leaf blocks
• isolation probe + strict-xfail wrapper (diagnostic; fails until phases 5-12)

Direct ExecutionEngine.init callers must now pass the server handler to keep the Roboflow error mappings; server error handlers also catch the workflows environment-value error.

Type of Change

• Refactoring (no functional changes)

Testing

All CI must be passing

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code where necessary, particularly in hard-to-understand areas
• My changes generate no new warnings or errors
• I have updated the documentation accordingly (if applicable)

Additional Context

N/A

[github-actions]

👋 Thanks for the pull request! Here is how automated Claude review works here, so you spend credits (and reviewer time) wisely.

🚦 This PR is marked Ready for review, so automated Claude review will run — and every pass spends real credits.

Warning

💸 The Claude reviewer bills in credits, not vibes

Automated review spins up a real agent that reads real code and spends real credits on every pass. It is glad to help — but it is not a rubber duck, a linter you poke in a loop, or a substitute for reading the contributing guide. Treat it like an expensive senior reviewer whose time you booked, and show up prepared.

Draft when unsure, Ready when you mean it:

• 🌱 Not sure the PR is in good shape yet? Keep it (or set it back) as a draft — drafts pause review, so you can push and iterate without burning credits on a moving target.
• 💪 Feel strong about the contents? Mark it Ready for review and the reviewer will take a look.

However you get there, arrive prepared:

• 🧱 Bring a SOLID, thorough PR. Point your local agent at our skills/ to tune it to our guidelines first — or, if you are one of those fabled carbon-based contributors, read them yourself. A half-baked diff costs exactly the same to review as a finished one.
• ✅ Resolve every comment before you re-request review. Re-requesting with threads still open means paying twice for the same conversation.
• 🔁 Do not use CI review as an inner loop for a local agent. The reviewer is not a step-by-step debugger — do the unfolding locally and arrive with the answer, not the search.
• 🙋 If something looks off, ask a human. One question to a maintainer is cheaper and faster than three rounds of agent re-review chasing a misread.

Reviews are not free. A draft costs nothing to review; a Ready PR is a promise that it is worth reviewing.

• Prefer to skip automated review entirely? Add the skip-claude-review label.

[github-actions]

🤖 Claude review started at commit 6090d3f1d4112bab3362588c0ce372c0f6e3768d.

New commits are not auto-reviewed. Add the claude-review label to request a re-review — the label is consumed when the review starts, so just add it again next time.

[grzegorz-roboflow]

Remaining 300+ imports are handled now, follow-up commit is coming

[github-actions]

Skills: review-workflows-execution-engine, review-workflows-blocks, review-http-api-server, review-core-infra, review-topic-test-hygiene, review-topic-backward-compat-and-versioning

No dedicated surface skill covers inference/core/interfaces/stream/inference_pipeline.py or inference/enterprise/workflows/** — generic review plus topic skills only.

Review scope. Large decontamination refactor (228 files). I traced the non-mechanical plumbing personally and ran two focused verification passes over the breadth-heavy areas:

• Step-error-handler move — all four production ExecutionEngine.init composition roots (http_api.py x2, inference_pipeline.py, inference_cli/.../local_image_adapter.py) now pass resolve_step_error_handler(); the relocated extended_roboflow_errors_handler is byte-identical to base, so HTTP/stream/CLI error mapping is preserved. The engine default flipping to mapping-free legacy only affects direct library callers that pass no handler — an intended decontamination contract.
• Enterprise blocks as a plugin — prepended into WORKFLOWS_PLUGINS, BLOCKS_SOURCE="workflows_core" preserves the init-parameter namespace and block ordering (core, enterprise, custom); no initializer collision; lazy import is strictly more robust than the removed top-level import.
• Logging — standard structlog ProcessorFormatter + foreign_pre_chain integration; parity test asserts correlation/execution/trace context survives the stdlib path for inference.core.workflows.* loggers.
• 30 model_manager removals — verified write-only dead state; __init__ and get_init_parameters() dropped in lockstep with no dangling references.
• Vendored helpers (LRUCache, image resize, cosine_similarity, ensure_parent_dir_exists) are faithful copies; protocols are annotation-only (not runtime_checkable). Lint baseline is internally consistent (375 rows = declared count); isolation probe is correctly strict-xfail / slow while decontamination is in progress.

Minor doubts (non-blocking): workflows/utils/text.py::str2bool now raises WorkflowsInvalidEnvironmentValueError(ValueError) instead of InvalidEnvironmentVariableError, and experimental emits a new warning category — both latent, no in-repo caller/filter depends on the old types today, and HTTP routes were updated to catch the new error. EXECUTION_ENGINE_V1_VERSION stays 1.15.1 with no user-facing behavior change (maintainer note: touches Execution Engine internals but is behavior-preserving for server users, so no version bump / changelog required).

Reviewed at HEAD: 6090d3f

[github-actions]

😎 PR passes the vibe-check and trust-me-bro verification.