Our sets of comparison SQL statements used for the unit tests had some

statements left with hard-coded index names, which aren't future-proof.
This change fixes those and prepares for the Django 1.7+ work by beefing

up how the test index name generation works. Each file of test SQL

statements now creates a local version of generate_index_name() using

the new make_generate_index_name() function, passing in the database

type. This allows generate_index_name() to be smarter about how it

handles certain types of indexes, particularly those for Postgres

(which, through Django 1.6, used a legacy set of rules for generating

index names).
There's also additional information now indicating if an index will be

used for Django's index_together.
Future changes will build upon this by creating variations of the index

name generation rules for Django 1.7, 1.8, and 1.9.
There's also new documentation in place for the affected functions, to

help with future maintenance.

Unit tests pass for SQLite3, MySQL, and Postgres.