feat(providers): add Cheaper Inference provider

[aiapienthusiast]

Summary

This PR adds Cheaper Inference as a model provider plugin.
It uses the same pattern as #200 (the OrcaRouter part).

Cheaper Inference is an OpenAI-compatible LLM gateway.
One API key gives access to models from several labs.
It also serves Anthropic Messages, so it works with claude_sdk and codex_sdk.

Usage:

export CHEAPER_INFERENCE_API_KEY=...
praxist setup --profile cheaperinference-api

Scope

• Affected modules or public contracts: new model_provider:cheaperinference plugin, cheaperinference-api setup profile, CHEAPER_INFERENCE_API_KEY credential.
• Compatibility considerations: additive only. Existing provider selection order does not change. The new key is checked after ORCAROUTER_API_KEY.
• Task-agnostic rationale, when shared Praxist behavior changes: no task-specific logic. Model ids pass through unchanged.

Verification

• uv run pytest tests/unit tests/conformance -q: pass, except 2 tests that also fail on main here (missing optional uvicorn and boto3).
• uv run pytest tests/hardening -q: pass.
• uv run ruff check praxist/ tests/ and uv run ruff format --check praxist/ tests/: pass.
• uv run pyrefly check: no new errors (existing missing-import errors in praxist/product_usage).
• uv run python scripts/build_docs_site.py --check-generated: pass.
• Live check: gpt-5.4-mini returned a reply through both provider paths.
• Not run: full coverage gate and real research tasks.

Checklist

• The change is focused and does not include unrelated generated files.
• Tests cover the affected behavior.
• Documentation, templates, examples, and skills were updated where needed.
• No credentials, private task data, or research-run artifacts are included.
• New dependencies and copied assets include their source and license terms.
• I have read .github/CONTRIBUTING.md and the contribution terms in LICENSE.md.

Details

Files:

• praxist/plugins/model_providers/cheaperinference/: plugin manifest and adapter. Default model gpt-5.4-mini.
• praxist/cli/_env.py, setup.py, start.py: key map, provider ref, base URL, setup profile, default provider and model.
• praxist/core/credentials.py, modeling.py, redaction.py: credential discovery, model id pass-through, ci_live_ key redaction.
• praxist/plugins/agent_runtimes/*: runtime allowlists, codex relay upstream and key var.
• praxist/plugins/workflow_stages/research_loop/*: provider env freeze, scoped env, default provider.
• Docs, skills, and docs/reference/cli.md (regenerated).

Tests: the same test files as #200, with Cheaper Inference cases.

Not copied from #200: the OpenRouter-style reasoning.effort relay override and the lossless context policy.
Those depend on provider behavior that maintainers may want to confirm first.

Live check: gpt-5.4-mini returned a reply through freeze_provider_env (Anthropic Messages, claude_sdk path).
It also returned a reply through the codex relay upstream (Chat Completions).