Fix exhibit TOC page numbering

[nonprofittechy]

Fixes #884.

Root cause

ALExhibitList._update_page_numbers() already includes every preceding exhibit cover in each exhibit's start_page. The TOC template then added loop.index, counting those covers a second time. That produced the reported increasing error: 3, 6, 9 instead of the physical content pages 3, 5, 7.

The no-cover expression also counted the default one-page TOC twice.

What changed

• Add ALExhibit.toc_page_number() as the single, documented calculation for the first uploaded page.
• Have the DOCX TOC template call that method for cover and no-cover output.
• Add regression tests for:
  • one cover per exhibit
  • cover pages disabled
  • a multi-page TOC

Before and after

Both screenshots come from the same local reproduction: three separate one-page exhibits assembled with a one-page TOC and one cover per exhibit.

Before: incorrect cumulative offset	After: physical start pages

Local validation

Installed each build into the localhost docassemble server with:

~/venv/bin/dainstall --server localhost .

Generated PDFs at http://localhost and extracted their TOC text:

• 3 one-page exhibits, covers enabled, 1-page TOC: 7-page PDF; starts 3, 5, 7
• 3 one-page exhibits, covers disabled, 1-page TOC: 4-page PDF; starts 2, 3, 4
• 28 one-page exhibits, covers enabled, 2-page TOC: 58-page PDF; starts 4, 6, …, 58

Additional checks:

• black --check docassemble/AssemblyLine/al_document.py docassemble/AssemblyLine/test_al_document.py
• changed-module mypy check
• git diff --check
• DOCX ZIP integrity check
• uv build --out-dir /tmp/issue884-dist

The repository-wide mypy run still reports four unrelated existing errors in language.py, al_general.py, and sessions.py; the changed module passes independently.

[nonprofittechy]

@rajeswari1301 this is not urgent at all (a very old correctness issue for a lightly used feature), but review if you can some day this week?