Define alias attribute methods in define_attribute_methods

[nvasilevski]

Fixes #48931

undefine_attribute_methods now removes alias attribute methods along with attribute methods. This commit changes define_attribute_methods to redefine methods back if any alias attributes were declared which provides applications and libraries an option to bring the alias methods back after using undefine_attribute_methods.

Implementation details

In addition to existing alias_attribute internal lists we will populate one more - aliases_by_attribute_name which will keep a list of alias attributes per attribute name. We will use this list in the define_attribute_methods to restore aliases any time the method is being redefined. Basically it makes define_attribute_methods responsible for defining both alias attribute methods and actual attribute methods for most cases apart from when a non-attribute thing is being aliased.

[byroot]

Some minor concern over that new attribute, but once this is cleared 👍

[byroot]

attribute_aliases is a class_attribute, so it's inherited and it makes sense to re-assign it.

However local_attribute_aliases and aliases_by_attribute_name are regular instance variables, so you can actually mutate them without issue.

Suggested change

	self.attribute_aliases = attribute_aliases.merge(new_name => old_name)
	self.local_attribute_aliases = local_attribute_aliases.merge(new_name => old_name)
	self.aliases_by_attribute_name[old_name] = aliases_by_attribute_name[old_name] + [new_name]
	self.attribute_aliases = attribute_aliases.merge(new_name => old_name)
	self.local_attribute_aliases[new_name] = old_name
	self.aliases_by_attribute_name[old_name] << new_name

[byroot]

Also I wonder if we really need to keep both. I see local_attribute_aliases is used here:

rails/activerecord/lib/active_record/attribute_methods.rb

Lines 74 to 76 in 9044d35

	local_attribute_aliases.each do |new_name, old_name|
	generate_alias_attribute_methods(code_generator, new_name, old_name)
	end

, but I think you could use aliases_by_attribute_name for that (or the opposite, what I'm trying to say is that it seem to me that we don't actually need both).