Add change nullability statement

[davidepaolotua]

This PR attempts to fix two things:

provide a way to change the nullability of a column in the migrator.

Related to #1039
Two new macros have been created:

• allow_nulls_for :column, which allows for the :column to receive null values
• forbid_nulls for :column, which set the not null constraint
  it is split from the change_type statement, so that you can theoretically do something like

change_type the_column : String
# migrate somehow the null fields present in the table
forbid_nulls_for :the_column

changes the order of the change_default statement with respect to the change_type statement

Currently, if you had something like:

change_type a : String
change_default a : String, default: "Whatever"

it would generate first the change of default, then the change of type, which is probably counterintuitive.
Now, it generates first the change of type, then the change of default

[jwoertink]

Nice! Let's give this a shot 🚀

[jwoertink]

@davidepaolotua I was just writing up some documentation and it just occurred to me.... why did we add allow_nulls_for and forbid_nulls when we have make_required and make_optional which generate the same SQL? 🤔

avram/src/avram/migrator/change_null_statement.cr

Lines 13 to 24 in bec9058

	def build
	if @required
	change = "SET"
	else
	change = "DROP"
	end
	String.build do |index|
	index << "ALTER TABLE #{@table}"
	index << " ALTER COLUMN #{@column}"
	index << " #{change} NOT NULL;"
	end

avram/src/avram/migrator/alter_table_statement.cr

Lines 104 to 106 in bec9058

	def build_nullability_statement(column_name, nullability)
	"ALTER TABLE #{@table_name} ALTER COLUMN #{column_name} #{nullability ? "DROP" : "SET"} NOT NULL;"
	end

I'm sure there was some logic behind having a second way to do this, but I'm not seeing why, and I don't remember 😅

[davidepaolotua]

Uhm... I wish I was as sure as you are...

imho I just completely missed the existence of that method - I prolly just looked in the alter_table file, and didn't occur to me that there was actually a separated file only for that case, while the other change_xxx are in alter table ^^'

we actually could remove those methods imho (just keeping the fix for the ordering of the ddl statements)

[jwoertink]

Ok. I'll at least open an issue on it. I'm sure I just forgot about the other methods too at the time 😂