Prepare to sync for ruby/ruby

[hsbt]

This changes are for syncing ruby/ruby repo.

• ruby/ruby repo needs test-unit instead of minitest.
• I reverted f5ea80d. Because this is for [win32/registry] Fallback to UTF-8 for unknown codepages ruby#7366. I don't have Arabic environment, but this issue is reported in 2023. Is it working only UTF8 locale? The current ruby/ruby implementation may worked at least.
• helper file is not recommended at ruby/ruby repo. I embedded it to test file.
• I'm not sure what's better to use volatile environment and (unused_value)[https://github.com/ruby/ruby/blob/master/test/win32/test_registry.rb#L86] block. This repository don't use that. What's do you think that?

@larskanis Can you look the above changes? After merging this, I enabled auto-sync feature this repository and ruby/ruby repository. You can develop this repository independently from ruby/ruby repo and we integrate these changes automatically for new Ruby release.

Unfortunately, we don't have sync feature for ruby/ruby to ruby/win32-registry. If we have those cases, I will create PRs manually.

[hsbt]

This change is testing now at ruby/ruby#11791

[larskanis]

• ruby/ruby repo needs test-unit instead of minitest.
• helper file is not recommended at ruby/ruby repo. I embedded it to test file.

Both is OK, of course. Thank you!

• I'm not sure what's better to use volatile environment and (unused_value)[https://github.com/ruby/ruby/blob/master/test/win32/test_registry.rb#L86] block. This repository don't use that. What's do you think that?

Using a random subkey versus a fixed one is probably necessary to run the tests in parallel. This is no problem in this repository, but in ruby/ruby. I can change this.

On the other hand I don't like having to clean up every test case per ensure path. This is tedious and should be done in a setup/teardown method IMHO.

• I reverted f5ea80d. Because this is for #7366 . I don't have Arabic environment, but this issue is reported in 2023. Is it working only UTF8 locale? The current ruby/ruby implementation may worked at least.

Both ruby/ruby#7366 and oneclick/rubyinstaller2#348 are fixed by just using UTF-8 as LOCALE. #7366 is outdated insofar, that I patched RubyInstaller to use UTF-8 always, since a year ago or so. So reverting it doesn't make sense to me.

There have been even more bug reports regarding non-UTF-8 registry keys, so that I decided to change it to consistent UTF-8 output. I only forgot to send the patch to ruby/ruby. But I created the win32-registry repo because of this inconsistent use of locale/UTF-8 string encoding and to have a way to properly run tests on the code. In contrast I didn't receive a single bug report due to changing the output to consistent UTF-8.

If f5ea80d is reverted, then the "test_utf8_encoding" test fails on non UTF-8 code pages. But not on GHA, since it is switched to 65001.

[larskanis]

I removed the reverted commit and merged this PR, so that I can check the parallel test and random registry path. Please raise your hand if the UTF-8 patch is still a concern for you!