FND-3849: add pyatlan_v9 creators for the 10 Tableau types

[cmgrote]

✨ Description

Adds creator() for the 10 Tableau asset types in pyatlan_v9: TableauSite, TableauProject, TableauWorkbook, TableauDatasource, TableauDashboard, TableauWorksheet, TableauDatasourceField, TableauCalculatedField, TableauWorksheetField, TableauDashboardField. Follows the Sigma (#1087) and Cognos (#1078) pattern: overlay + parity tests + integration test.

TableauDatasource.creator(
    name="Orders",
    project_qualified_name="default/tableau/1712345678/site-guid/project-guid",
    datasource_id="ds-guid",
    workbook_qualified_name="default/tableau/1712345678/site-guid/project-guid/wb-guid",
)
# embedded: qualified_name == ".../project-guid/wb-guid/ds-guid", is_published=False

• Exact app parity: each creator emits the qualifiedName atlan-tableau-app emits today (FND-3838 census, FND-3849 analysis). Every census variant collapses to one shape per type.
• Parent qualifiedName in, ancestors derived: creators take the parent's qualifiedName plus the source id; site / project / workbook / connection qualifiedNames are sliced from it. A parent at the wrong depth raises ValueError; blank inputs raise (validate_required_fields).
• Projects stay flat ({site}/{project}) even when nested. parent_project_qualified_name sets parent_project, is_top_level_project and top_level_project_qualified_name (defaults to the parent).
• Embedded vs published datasources: passing workbook_qualified_name gives the embedded shape and sets the workbook relationship; the project relationship is always set. Datasource and calculated fields infer published vs embedded from the datasource qualifiedName depth.
• Worksheet / dashboard fields set the tableau_-prefixed denormalised attributes and the tableau_worksheet / tableau_dashboard relationship, as the app does.
• top_level_project_qualified_name defaults to the immediate project (correct when the project hierarchy is empty); callers with deeper nesting pass it. projectHierarchy stays caller-set, since it needs ancestor names.
• Overlays in pyatlan_v9/model/assets/_overlays/tableau_*.py; only the 10 Tableau classes are taken from the regeneration (pkl sdkOnly=python with these overlays, same flags as sync-python-sdk.yaml).

Linear: FND-3849 (analysis table in the issue comments)
Paired models PR: https://github.com/atlanhq/models/pull/2088 (adds the matching qualifiedNameFormula; only TableauSite's generated pattern changes).

Merge order: this PR first, then models#2088. The sync bot regenerates with overlays from atlan-python main on each models master push, so in this order its sync is a no-op for Tableau; the other way round it regenerates the classes without the creators.

---

🧩 Type of change

• 🚀 New feature (non-breaking change that adds functionality)

---

✅ How has this been tested?

• tests_v9/unit/model/tableau_<type>_test.py (10 files, 125 tests): literal app qualifiedName per type (published and embedded where applicable), pattern match, validate(for_creation=True), parent relationship and denormalised attrs, missing / blank inputs, malformed parent qualifiedNames, cross-site top-level project, create alias, updater / trim_to_required.
• tests_v9/integration/tableau_asset_test.py (12 tests): builds site → top-level project → nested project → workbook → published and embedded datasources, dashboard, worksheet and every field type through the creators, retrieves each and asserts qualifiedName, parent link and denormalised attrs, then purges. Passes locally against a dev tenant.
• Generator fidelity: models#2088 + these overlays reproduces this PR's 10 classes with no diff.
• pytest tests_v9/unit: 8152 passed, 5 skipped. pre-commit clean.

---

📋 Checklist

• My code follows the project’s style guidelines
• I’ve performed a self-review of my code
• There are no new warnings from my changes
• I’ve added tests to cover my changes
• All new and existing tests pass locally

🤖 Generated with Claude Code

https://claude.ai/code/session_016K2wFCg1G6Hxv8j2TKW7cB

[linear]

FND-3849