feat(tables): row_count/column_count/dimensions + sizing round-trip lock — Tables 2.0 Phase 4 (closes epic)

[MHoroszowski]

Phase 4 of issue #12 (Tables 2.0): closing PR — sizing & ergonomics

Closes the Tables 2.0 epic. Phase 1 (PR #37) shipped row/column add/remove. Phase 2 (PR #40) shipped the table style API. Phase 3 (PR #41) shipped range-style merge_cells/split_cells with Cell inspection accessors. This PR adds the last two unchecked sub-features on issue #12 and closes the issue entirely.

Public surface added (additive — no existing API changed)

table.row_count       # int — equivalent to len(table.rows), no _RowCollection instantiation
table.column_count    # int — equivalent to len(table.columns), no _ColumnCollection instantiation
table.dimensions      # (row_count, column_count) tuple — symmetrical with TcRange.dimensions

All three are read-only (AttributeError on assignment). They're already reactive: table.rows.add() increments row_count, table.columns.remove(0) decrements column_count, etc.

Round-trip regression-lock for sizing

_Row.height and _Column.width setters already wrote to <a:tr>/@h and <a:gridCol>/@w correctly, and a smoke test confirmed round-trip through save+reload already worked on master. This PR pins that down with five regression tests covering single-row height, single-column width, mixed-height row sets, mixed-width column sets, and count-property preservation through round-trip — so any future change that breaks EMU preservation fails CI immediately.

Out of scope (deferred to follow-ups, not omitted from issue #12)

• NumPy-style table[r, c] indexing — table.cell(r, c) is canonical; adding __getitem__ would conflict with _RowCollection.__getitem__ on table.rows[i] and confuse the data model.
• Bulk getter helpers (row_heights / column_widths returning lists) — list-comprehension is one line; no API value-add.
• Constraint or auto-fit logic — PowerPoint's job at render time.

Tests

• 22 new pytest cases in tests/test_tables_phase4.py covering each new count property (including reactivity to .rows.add() / .columns.remove() etc.), the read-only setter checks, the round-trip preservation matrix, and Phase-1/2/3 regression checks. Full suite: 3426 passed.
• 7 new behave scenarios in features/tbl-sizing.feature: row_count read, column_count read, dimensions read, count-after-add, count-after-remove, height round-trip via stream, width round-trip via stream. Full behave: 1036 scenarios passed, 0 failed (baseline 1029 + 7 new).
• Ruff clean: ruff check src tests → All checks passed; ruff format --check → no diff.

Reporting contract (CLAUDE.md §7)

$ python3 -m pytest tests/ -q | tail -3
........................................................................ [ 98%]
..........................................                               [100%]
3426 passed in 4.79s

$ python3 -m ruff check src tests | tail -3
All checks passed!

$ python3 -m behave features/ --no-color 2>&1 | tail -3
1036 scenarios passed, 0 failed, 0 skipped
3120 steps passed, 0 failed, 0 skipped
Took 0min 1.603s

UAT

• uat_tables_phase4.py (untracked per CLAUDE.md §6) at repo root.
• Builds two tables: a 4×4 with explicit graduated row heights (1.31" → 0.55") and monotonically widening column widths (0.98" → 3.39"), and a 3×4 grown into a 4×4 visually-square grid via the new count properties driving the resize loop.
• Surfaces a real PowerPoint quirk worth documenting: PowerPoint enforces a content-driven minimum row height (~0.4" for default padding + font line height). Row heights below that floor are visually clamped at render time, but the EMU value still round-trips byte-stable in the XML — meaning future XML probes / automation pipelines see your exact value even when PowerPoint paints something taller.
• Visual UAT signed off by maintainer.

Tables 2.0 epic accounting (all eight sub-features now shipped)

#	Sub-feature	Phase	PR
1	Table.apply_style(name_or_id)	Phase 2	#40
2	Table.rows.add() / .remove()	Phase 1	#37
3	Table.columns.add() / .remove()	Phase 1	#37
4	first_row / banded_rows / last_row etc. bound to a real tableStyleId	Phase 2	#40 (toggles already existed; now have public style API)
5	Table.merge_cells / split_cells (idempotent)	Phase 3	#41
6	Cell.gridSpan / rowSpan / hMerge / vMerge read-only	Phase 3	#41 (snake_case on the Python API)
7	Per-row height / per-column width round-trip	Phase 4	this PR (regression-locked)
8	Table.row_count / column_count ergonomics	Phase 4	this PR

Closes #12