Fix spec for trap_thread_dump_signal

[eregon]

Description

• The spec was timing-dependant and therefore flaky.
• Use a common pattern guaranteed to match.
• Always wait for the signal since Process.kill can be asynchronous (and in real cases always is).
• See Process.kill does not call the registered handler immediately when called for the current process on the main Thread truffleruby/truffleruby#4383 (comment)

Locally I added a 1000.times { ... } around to spec and that way I could repro the failure locally (100 was not enough).
Before the fix it failed reliably a few times, after it passes reliably.

Type of change

Please delete options that are not relevant.

• spec fix

Checklist:

Your PR is ready for review once the following checklist is
complete. You can also add some checks if you want to.

• Tests have been added for any changes to behaviour of the code
• New and existing tests are passing locally and on CI
• bundle exec rubocop reports no offenses
• RDoc comments have been updated
• CHANGELOG.md has been updated (not user-facing so no entry)

[luke-hill]

@eregon think there's a JRuby error hre

[eregon]

Ah yes because JRuby runs the handler on a non-main Thread, I'll adjust the pattern for that.

[eregon]

@luke-hill Done, please start the CI and take another look.
I ran the spec locally on JRuby and it passed now.