Prepare library API and docs for PEtab v2 problems

[dweindl]

Closes #327.

Prepares benchmark_models_petab and the repo docs so that problems/<id>/v2/ directories can be added later without further groundwork. No v2/ problem directory is created here — that's separate, later work.

• Splits library access into benchmark_models_petab.v1 / .v2 submodules, mirroring petab.v1/petab.v2. Each exposes get_problem, get_problem_yaml_path, get_simulation_df. Top-level get_problem etc. keep their current (v1) behavior unchanged (base.py/test_base.py renamed to v1.py/test_v1.py).
• v2.get_problem_yaml_path/get_problem raise a clear error when a problem has no v2/ yet (the normal case for essentially every problem right now).
• Documents the v1/v2 equivalence policy in CONTRIBUTING.md: a problem's v1/ and v2/ encodings must represent exactly the same mathematical problem (same model, data, objective), though v2/ files may look structurally different to make idiomatic use of v2 features.
• Adds an optional PR-template checklist item pointing at that policy.
• Both v1/ and v2/ are documented as optional (at least one must be present), rather than implying v1/ is mandatory.

Out of scope, left for follow-up work: generating/hand-curating any actual problems/<id>/v2/ directory, an automated v1-vs-v2 equivalence checker, and updating the README overview table / INSTALL.md usage docs for v2 (postponed until real v2 problems exist).

🤖 Generated with Claude Code

[dweindl]

This will probably need some refinement. How do we want to deal with the whole parameter scaling topic? Encode it in the model or condition table, ignore it, ...?