[Confluence] Fix pagination for get_all_* methods and unify _get_paged across Cloud/Server

[Zircoz]

Summary

Fixes #1598
Fixes #1480

Multiple get_all_* methods (get_all_pages_from_space, get_all_pages_from_space_trash, get_all_draft_pages_from_space, etc.) only returned the first page of results because they called self.get() directly instead of self._get_paged(). Additionally, pagination broke when the Confluence API returned _links.next as a plain URL string rather than a {"href": "..."} dict.

Changes

Pagination fix (core issue)

• Switch 10 get_all_* methods to use _get_paged on both Server and Cloud, so they now return fully-paginated generators instead of single-page dicts
• Unify _get_paged into ConfluenceBase — removed duplicate implementations from ConfluenceCloudBase and ConfluenceServerBase, replacing them with a single method in the shared base class that handles both str and dict formats for _links.next
• Fix relative pagination URLs — when the API returns a relative path (e.g. /rest/api/content?cursor=1), the base URL is now correctly prepended by stripping the api_root suffix, preserving Cloud API gateway routing prefixes

Cloud URL/routing fixes

• Fix Cloud api_root from wiki/api/v2 to wiki/rest/api — the previous default was incorrect: every endpoint in the Cloud class uses v1 REST API paths (content/{id}, space/{id}, content/search?cql=, etc.), but they were being prefixed with the v2 base path (wiki/api/v2). The v2 API has an entirely different resource model (pages/{id}, spaces/{id}, no CQL search) so the old configuration produced invalid URLs. This is a bugfix, not a v2→v1 downgrade.
• Recognize api.atlassian.com in Cloud detection — the Confluence wrapper now correctly routes OAuth2 API gateway URLs to ConfluenceCloud
• Support explicit cloud= kwarg — allows users to force Cloud or Server routing regardless of URL heuristics

Tests

• New test_confluence_routing.py — covers .atlassian.net, .jira.com, api.atlassian.com, self-hosted URLs, and explicit cloud=True/False override
• Pagination edge-case tests for both Cloud and Server: string next link, dict next link, missing href, None next link, no results key, and relative next link resolution
• Updated existing tests to expect generator return values from get_all_* methods

Breaking change

get_all_pages_from_space, get_all_blog_posts_from_space, get_all_pages_by_label, get_all_blog_posts_by_label, get_all_draft_pages_from_space, get_all_draft_blog_posts_from_space, get_trash_content, get_all_pages_from_space_trash, and get_all_blog_posts_from_space_trash now return generators (from _get_paged) instead of dicts. Callers that accessed result["results"] must switch to iterating the return value directly (e.g. list(result)).

Test plan

• pytest tests/confluence/ — all new and updated tests pass
• Verify get_all_pages_from_space returns all pages (not just first 25) on a Confluence Server instance with >25 pages
• Verify the same on Confluence Cloud
• Verify Cloud OAuth2 gateway URLs (api.atlassian.com/ex/confluence/...) route and paginate correctly

🤖 Generated with Claude Code