Reformat LaTeX source of tables for better parsing and diffing

[jcrippen]

The LaTeX source code currently represents each row of a table as a single row of text. This makes it easy to read the source, but it actually complicates some automatic tooling. Even if only a single character is changed in one cell, diff considers the entire row to be different. Since the rows are very long this makes interpreting diffs – and thus git output – very hard. It also makes the data harder to parse since a program has to know quite a bit about LaTeX table syntax to pick apart the individual cells in each row.

An alternative format is to put each cell on its own line in the LaTeX source. A line like

cell 0 & cell 1 & cell 2\\

would instead be something like

cell 0
      & cell 1
              & cell 2\\

So that each cell is on a separate line. A change made to a single cell would be seen in a diff by the change of a single line. Parsing a row would be reading each line up to one ending with \\ and then filtering the LaTeX cruft like &. Since each cell is on its own line, the count of lines in a block would track the number of cells in a row and thus make indexing simpler.