Test maintenance improvements from #759

[danielshahaf]

Spun off from @m0vie's #759:

• [ ] In the output of a failing test, where the actual and expected values are tabulated, add a marker on the left margin of lines that failed comparison.
• [ ] In the output of a failing test, where the actual and expected values are tabulated, in the "actual" column, show both the postprocessed expected_region_highlight-style values (1-indexed, inclusive of both ends) for eyeballability and the raw region_highlight-style values (0-indexed, half-open) that were verified.
• [ ] In the output of a failing test, on the line of a failing test point, the "observed" datum is given in expected_region_highlight-style values, even though observations are in region_highlight-style values.
• [ ] Ensure the two-formats issue is clearly documented in tests/README.md.
• [ ] Ensure the existence of tests/generate.zsh is clearly documented in tests/README.md.
• [ ] (Likewise any other scripts in tests/.)
• [ ] Rename expected_region_highlight to a name that reflects the data format differences to region_highlight (Hungarian notation)