fix: avoid panic when attempting to use test server across fork

[chris-olszewski]

What was changed

Avoid dropping the tokio runtime if we're in a fork.

Why?

We document that objects that hold this runtime handle cannot be used across threads and throw specific errors explaining this if it is attempted

sdk-ruby/temporalio/test/client_test.rb

Lines 155 to 167 in 1cda7a1

	def test_fork
	# Cannot use client on other side of fork from where created
	pre_fork_client = env.client
	pid = fork do
	pre_fork_client.start_workflow(
	'some-workflow', id: "wf-#{SecureRandom.uuid}", task_queue: "tq-#{SecureRandom.uuid}"
	)
	rescue Temporalio::Internal::Bridge::Error => e
	exit! 123 if e.message.start_with?('Cannot use client across forks')
	raise
	end
	_, status = Process.wait2(pid)
	assert_equal 123, status.exitstatus

The issue is that these useful errors can be hidden by a panic from dropping the runtime handle which allows the expected errors to surface to users.

Checklist

1. Closes [Bug] Parallel test configuration in Rails causes crash #407

2. How was this tested:
   Initial commit added failing test. Tested with Configure parallel tests in rails app (which doesn't work - yet) samples-ruby#78 (review) and verified that the errors explaining invalid usage across forks are now displayed instead of a panic.

3. Any docs updates needed?
   N/A