Change operations v1.1: table ops, positioned node insert, change grammar

[ronaldtse]

Change operations v1.1

The change tier is extended so every construct in the model is
addressable for collaborative editing (OT/CRDT-grade). This answers
two committee questions on the document:

1. How are fine-grained operations on tables and computed content
   exchanged? — the whole change family now covers row/column
   insert/delete/move, cell update, cell merge, and cell split with
   grid coordinates that account for colspan/rowspan. Computed cells
   (formulas in cells) ride the existing stem construct with an open
   notation key — no new types needed.

2. How does change ordering relate to concurrent editing? —
   the existing parentIdentifier field already provides the causal
   DAG of changes. The new clause 11 v1.1 subclauses (in the spec
   PR) state plainly: changes are operation records, ordering is
   the DAG, and how concurrent operations transform against each
   other (OT, CRDT, diff/patch) is out of scope — the same neutrality
   pattern as serialization.

Models

• models/change/tables/ — new family:
  • TableChange root
  • TableArea enum (head / body / foot)
  • CellCoordinate (row, column — zero-based grid slots)
  • RowInsert / RowDelete / RowMove (with area)
  • ColumnInsert / ColumnDelete / ColumnMove (with template
    cell)
  • CellUpdate / CellMerge / CellSplit
• models/change/NodePosition.lml — new: container reference +
  zero-based child ordinal.
• models/change/NodeInsert.lml — gains +position (was: only
  +content; insertion point is now explicit).

Grammar

• grammars/basicdoc.rnc — change grammar (changeset + change +
  all operations) appended in lockstep with the LML. Targets use
  plain text (cross-document identifiers), not IDREF.
• grammars/basicdoc-compile.rnc — start = document | changeset.

Fixtures

• examples/table-changes.{xml,yaml} — twin ChangeSet exercising
  every operation over the line-item table (li1t1) in
  document.xml, including a computed cell (FormulaBlock inside a
  ParagraphTableCell): row insert → column insert → header cell
  update → computed cell update → merge → split → row move
  (body→foot) → column move → row delete → positioned node insert
  → node delete (with hash) → node move → cursor-range text edit →
  register entry update.

Views

• views/Changes.lml — extended with the new family and the
  inheritance associations.
• views/TableChanges.lml — new focused plate.
• New PNGs: images/Changes.png (regenerated), images/TableChanges.png.

Gates

rake check   ✓ (lint, parity, verify, render)
rake fixtures ✓ (XML + YAML, all four files)
rake profiles ✓ (legalexchange/1.0)

135 model files, 17 views, 135 types.

What did NOT change

• The content tiers themselves (document/sections/blocks/inline) are
  unchanged. The audit confirmed every content construct is already
  one of four addressable shapes (string content, key/value slot,
  ordered sequence, table grid); the missing operations on
  ordered sequences (positioned node insert) and tables (the whole
  family) are exactly what this PR adds.

The spec half (clause 11 subclauses, Annex E business interchange,
Annex D extension, submodule pin) follows in the cc-lightweight-doc
PR.