Preserve 'linked' sideload data in PaginatedList (fixes #709)

[dylangriffinshub]

Summary

Fixes #709. Compound-document Canvas responses (e.g. the grade change log) return a linked section with sideloaded users/courses/assignments. PaginatedList discarded this data whenever _root was set, so callers lost the objects needed to resolve identifiers such as sis_user_id without making extra API calls.

Changes

• Collect the linked section from the original response before any _root extraction.
• Merge sideloaded objects across every page, deduplicating by id.
• Expose the collected data through a read-only linked property on PaginatedList.
• Add fixtures and tests for both single-page and multi-page compound documents.

Example

```python
events = course.get_grade_change_events()
for event in events:
...

sideloaded users/courses/assignments are now available:

events.linked["users"] # -> [{"id": "U1", "name": "Alice"}, ...]
events.linked["courses"] # -> [{"id": "C1", "name": "Math"}]
```

Backward compatible: responses without a linked section return an empty dict. All existing test_paginated_list / test_grade_change_log tests still pass, plus 2 new tests.

Closes #709