Convert all tables to use UTF8MB4 charset

We have seen an increasing number of errors recently relating to publishers attempting to use 4 byte characters in Whitehall.
This commit adds a data migration which will convert all of the tables to use the UTF8MB4 charset.
This uses a data migration because it is not subject to the 15 minute limit for schema migrations applied in the kubernetes configuration.

The migration will need to be run out of hours as the tables will need to be locked during conversion.

config/database.yml

@@ -1,5 +1,5 @@
 default: &default
-  encoding: utf8
+  encoding: utf8mb4
   adapter: mysql2
   prepared_statements: true
   variables:

db/data_migration/2024121711524600_convert_text_columns_to_utf8mb4.rb

@@ -0,0 +1,13 @@
+# Convert all tables to utf8mb4 in order to improve support for non-English languages.
+
+connection = ActiveRecord::Base.connection
+# Disable foreign key constraints because foreign keys using strings will prevent conversion due to a charset mismatch.
+connection.execute "SET foreign_key_checks = 0;"
+connection.tables.each do |table|
+  next if table == "schema_migrations"
+
+  puts "START: Converting table #{table} to utf8mb4"
+  connection.execute "ALTER TABLE `#{table}` CONVERT TO CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;"
+  puts "END: Converted table #{table} to utf8mb4"
+end
+connection.execute "SET foreign_key_checks = 1;"