✅ Fix JRuby local backtrace test assertions

[nevans]

A bug was fixed in jruby-head, but that may not be in current releases:

• JRuby Issue: raise should re-populate exception trace set to nil jruby/jruby#9528
• Fixed by: raise should re-populate exception trace set to nil jruby/jruby#9528

With that issue fixed, these tests don't need to be marked pending! 😄

BUT, JRuby does still have some incongruity between caller(1) and raise rescue $!.backtrace[1..]. Some ruby block stack frames in caller are replaced by java stack frames in Exception#backtrace. For example:

--- Kernel#caller
+++ Exception#backtrace
  /home/nick/.local/share/rubies/jruby-dev/lib/ruby/gems/shared/gems/test-unit-3.7.8/lib/test/unit/testcase.rb:632:in 'block in run'
- /home/nick/.local/share/rubies/jruby-dev/lib/ruby/gems/shared/gems/test-unit-3.7.8/lib/test/unit/testcase.rb:631:in 'catch'
+ org/jruby/RubyKernel.java:1604:in 'catch'
+ org/jruby/RubyKernel.java:1599:in 'catch'
  /home/nick/.local/share/rubies/jruby-dev/lib/ruby/gems/shared/gems/test-unit-3.7.8/lib/test/unit/testcase.rb:631:in 'run'

The workaround is relatively simple: use a locally generated exception to generate the stack frames for comparison.

[nevans]

@headius I found this curious incongruity last week: caller(1..) == (raise rescue $!.backtrace[1..]) isn't always true for JRuby. It's not a big deal, so I didn't make a JRuby issue for it.

But caller locations and exception backtrace locations seem to always agree under CRuby and TruffleRuby. Do you think it's worth making a ruby spec update?