Only skip specific failures on JRuby instead of ignoring the whole CI feedback

[deivid-rodriguez]

This PR is just some experiment to proof that jruby/ruby-maven#2 fixes:

• Issue reported at JRuby tests failing #479 (comment), where jruby tests fail to even get started in CI (bundle install fails).
• unsupported Java version "16", defaulting to 1.7 [ERROR] Internal error... #520.
• unsupported Java version "16", defaulting to 1.7 jruby/ruby-maven#3.
• Bundler fails on rubocop-rspec with JRuby rubygems/rubygems#4685.

[deivid-rodriguez]

This PR is just to proof that the changes work, but it should not be merged. Once jruby/ruby-maven#2 is merged and released, things should start working without any changes.

[deivid-rodriguez]

Alright so dependent PRs are moving forward, and this is now only missing a ruby-maven release.

However, the gem became uninstallable again on jruby after #525, since gemified stringio does not yet support jruby (WIP for that is ruby/stringio#21).

[headius]

This would be worth revisiting with latest SnakeYAML. On master, I now see 5F7E. Some of these may be valid differences (SnakeYAML appears to report line/column positioning a bit differently) but several appear to be fixable errors or bugs. We also addressed the stringio gemification for JRuby 9.4.

[deivid-rodriguez]

@headius I completely lost context of this PR. I think the original issues I wanted to address are now fixed, but a few others remain. I changed the scope of this PR to:

• Use Bundler to install dependencies and run tests.
• Only skip the particular tests that are failing instead of ignoring the whole test run outcome. I think this should help preventing new regressions.

Let's see what CI says.

[deivid-rodriguez]

Except for the specs that are marked as pending, everything is now green on JRuby, so I set this as ready for review.

[headius]

@deivid-rodriguez Very nice! Good idea to skip some tests that are problematic... some might not be fixable due to minor differences in SnakeYAML (positioning information etc).

@tenderlove Any objections to merging this PR? It marks a number of tests as pending, but I'm not sure they are fixable and it would be good to have a green-ish run so we can see when things regress.

[hsbt]

@deivid-rodriguez 👍 Thanks!