Integrate various fixes around finding model classes, including https://github.com/ctran/annotate_models/pull/82 and https://github.com/ctran/annotate_models/pull/85

By default, ObjectSpace.each_object is disabled in JRuby. However,
ObjectSpace.each_object(Class) still works.

This reverts commit 8d3c00c7.

Sorry, but database order should be the default unless a user really wants it alphabetized. (We should also allow a .annotate file so users can set their own preferred options more easily.)

Conflicts:

	.gitignore
	bin/annotate
	lib/annotate/annotate_models.rb
	lib/tasks/annotate_models.rake

Conflicts:

	lib/annotate/annotate_models.rb

This approach eliminates necessity to guess model name by file name that
allow to process models with any kind of non-standard capitalization in
class and module names. Though, it still requires that underscored file
name matches underscored model class name. If file is located not at the
top level of models directory then we assume that model is either nested
in the corresponding chain of modules or is not nested at all.

Annotation position defaulted to 'before'. Bugfix: annotate_models.rake uses string instead of symbol.

Allow task loading from Rakefile for gems (plugin installation already auto-detects). Add skip_on_db_migrate option as well for people that don't want it.

Fix options parsing to convert strings to proper booleans. Change annotate to use options hash instead of ENV.

Add a generator (annotate_models:install) to install a Rakefile that automatically annotates models when db:migrate is invoked.

We had stripped the trailing newlines from our Schema Information section. This small fix supports that case.