CI should test against major downstream consumers

[flavorjones]

Nokogiri is a major part of the Ruby and Rails ecosystem, and we want to make sure we're doing our best to avoid introducing surprising breaking changes in downstream consumers by testing the latest (and maybe HEAD) of some downstream libraries against Nokogiri's origin/main.

I wrote a script to find transitive reverse dependencies with lots of downloads, but there are other consumers like ActiveSupport::XmlMini that we should also consider.

As a first pass, I think the downstream consumers that we test should include gems likely to depend on edge case behavior (like semi-public APIs, broken markup cleanup, or sanitization):

• nokogumbo (we already do this as part of testing gem installation, see scripts/test-nokogumbo-compatibility)
• loofah ci: test downstream libraries against this version of nokogiri #2535
• sanitize ci: test downstream libraries against this version of nokogiri #2535
• rails-html-sanitizer ci: test downstream libraries against this version of nokogiri #2535
• ActiveSupport::XmlMini
• signer (Not compatible with Nokogiri >= 1.12.0  ebeigarts/signer#30)
• Viewpoint ([bug] Possible regression in XML generation #2317)
• webrat? (Breaking Change in to_xpath method breaks webrat #2469)

Here's the output from the script linked to above, all transitive reverse dependencies with more than 50 million downloads:

• mime-types: 427192129
• actionpack: 301924540
• rails: 291427727
• actionmailer: 286621408
• railties: 285189135
• sprockets: 262728161
• mime-types-data: 246581469
• actionview: 245752445
• loofah: 229877006
• sprockets-rails: 220115681
• rails-html-sanitizer: 214929483
• rails-dom-testing: 214021258
• rest-client: 197506600
• jquery-rails: 164304611
• coffee-rails: 149628440
• httparty: 140741625
• actioncable: 136061936
• awesome_print: 126815236
• rspec-rails: 125020709
• mimemagic: 118289077
• capybara: 115351581
• responders: 108551246
• xpath: 105908875
• devise: 102297652
• kaminari: 85638960
• activestorage: 85143954
• haml: 83777188
• fog-core: 80990363
• fog-json: 76526485
• marcel: 76479842
• fog-aws: 73467593
• fog-xml: 70598608
• autoprefixer-rails: 66030405
• rspec_junit_formatter: 63620797
• factory_bot_rails: 60680591
• oauth: 59455942
• paperclip: 53141734
• kaminari-actionview: 52880566
• web-console: 51907828
• fastlane: 51243313
• lograge: 50700055

[flavorjones]

I've labeled this "help wanted" because I think it's something that doesn't require deep knowledge of Nokogiri's internals to do.

[flavorjones]

Another example that would have been helpful to have is webrat, see #2469

Edit: the webrat test suite is suffering from bitrot pretty badly and I couldn't easily get it running.

[flavorjones]

Done! See https://github.com/sparklemotion/nokogiri/actions/runs/2294173903