fix: register anchor collector at priority -1 to catch extension-generated anchors - #4161

Fix #3690

Problem

The _ExtractAnchorsTreeprocessor is registered at priority 5, but PyMdown Tabbed's TabbedTreeprocessor runs at priority 4. Since Python-Markdown's Registry sorts by priority descending, priority 5 runs before priority 4, causing the anchor collector to miss slugified tab IDs generated by the Tabbed extension.

When Tabbed is configured with a slugify function, readable IDs like my-tab-title or headline-tab-a (with combine_header_slug) are added to <input> elements during the treeprocessor phase -- after the anchor collector has already scanned the tree.

Any link like [see instructions](page.md#my-tab-title) then produces a false-positive anchor validation warning, even though the anchor exists in the final HTML.

Fix

Changed the registration priority of _ExtractAnchorsTreeprocessor from 5 to -1, ensuring it runs after all other treeprocessors (including TabbedTreeprocessor at priority 4 and _RelativePathTreeprocessor at priority 0) so it captures all dynamically generated anchors.

Verification

• All 130 structure tests pass
• All anchor-related build tests pass (test_anchor_no_warning, test_anchor_warning, etc.)
• Manual test with pymdownx.tabbed confirms tab-a, tab-b, and headline-tab-a/headline-tab-b (with combine_header_slug) are all correctly collected