Return all muscles on the first API page by default

[cpruijsen]

Title: Return all muscles on the first API page by default

# Proposed Changes

- `MuscleViewSet` now uses `MuscleLimitOffsetPagination` (in `wger/utils/pagination.py`), which sets `default_limit = max_limit` (999). A request without `limit` gets all muscles on the first page.
- Why: the muscle dropdown in the exercise form fetches `/api/v2/muscle/` without a `limit` and reads only the first page's `results`. With `PAGE_SIZE` at 20, an instance with more than 20 muscles (the default data has 16) loses some from the dropdown.
- The response envelope (`count`, `next`, `previous`, `results`) is unchanged, and an explicit `limit`/`offset` still paginates.
- New test in `MuscleApiTestCase` with 36 muscles. It fails on master with `AssertionError: 20 != 36`.

`pagination_class = None` would return a bare list and break the web client, which maps `data.results`. The alternative is to fix `getMuscles` in `wger-project/react` to pass `limit` or follow `next`. I can do that instead, or as well, if you prefer.

Not changed here:

- `getEquipment` in the React client has the same pattern. The same one-line `pagination_class` would cover equipment and categories if you want it in this PR.
- The muscle endpoint is wrapped in `cache_page(EXERCISE_CACHE_TTL)`, and nothing clears that cache when a muscle is added, so a new muscle can be missing until the cache expires. That may be the "caching" in the first comment. It needs an invalidation hook, so this PR does not close the issue.

## Related Issue(s)

See #1737 (fixes the 20-muscle cap only)

## Please check that the PR fulfills these requirements

- [x] Tests for the changes have been added (for bug fixes / features)
- [x] Code has been formatted to avoid unnecessary diffs (`ruff format && ruff check --fix`)
- [ ] If the feature is big enough or if there are manual steps needed (deployment changes etc.), write a small writeup in `CHANGELOG.md`