fix: map WRAP_ACTIONS in the ModelExtension binding

[wyongzhi]

What was the problem/requirement? (What/Why)

PyModelExtension was missing the WRAP_ACTIONS member, so the mandatory catch-all arm in From<ModelExtension> for PyModelExtension silently reported it as EXPR. A ModelProfile built with WRAP_ACTIONS therefore enabled EXPR instead, and templates using onWrapEnvEnter/onWrapTaskRun/onWrapEnvExit were rejected with an error that did not name the real cause.

Minimal repro on 0.11.1:

from openjd.model._v1.types import ModelExtension
assert ModelExtension.from_str("WRAP_ACTIONS") is ModelExtension.from_str("EXPR")  # holds, should not

Why it was missed: WrapActions landed in the Rust crate after this binding's enum was written. #285, #306, and #318 all touched WRAP_ACTIONS support in the template layer (template_types.rs, _model.py), but none touched profile.rs, where the extension-enablement enum lives. Because upstream ModelExtension is #[non_exhaustive], the catch-all arm is compiler-mandated, so the missing arm produced no warning.

What was the solution? (How)

• Add WRAP_ACTIONS = 4 to PyModelExtension and both From conversion arms
• Replace the catch-all's comment with a warning that it misrepresents unmapped variants as EXPR, and document the three edits a new upstream variant needs
• Update the .pyi stub
• Add TestModelExtension (drift guard): asserts the binding's member set equals openjd.model.v2023_09.ExtensionName, and that every spec name round-trips to its own member via from_str. Either test fails on the next unmapped variant. This is the guard the compiler cannot provide under #[non_exhaustive].
• Add WRAP_ACTIONS to the pickle round-trip parametrize list

What is the impact of this change?

ModelProfile built with WRAP_ACTIONS (directly, via from_strings, or from a template's extensions: list) now actually enables WRAP_ACTIONS. This unblocks downstream consumers that declare supported_extensions from ExtensionName, for which WRAP_ACTIONS was silently never enabled.

How was this change tested?

• Have you run the unit tests? Yes — full suite: 5453 passed, 24 skipped, 3 pre-existing xfailed; the 94% coverage gate passed
• test_name_round_trips_to_its_own_member[WRAP_ACTIONS] fails without the fix and passes with it
• ruff, black, mypy clean

Was this change documented?

• Are relevant docstrings in the code base updated? Yes — the catch-all arm's comment now documents the misrepresentation hazard and the three edits required when upstream gains a variant, and points at the drift-guard test. The .pyi stub lists the new member.

Is this a breaking change?

No. It adds an enum member and corrects a mapping; no existing public contract changes. Code that (incorrectly) relied on from_str("WRAP_ACTIONS") returning EXPR would change behavior, but that behavior was the defect.

Does this change impact security?

No. No new files, directories, or privilege boundaries; no threat modeling needed.

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.