Move screenshot clearing to be before error raise

[TaylorBrysonRouse]

When a spec or test would fail from a difference, the following specs/tests would fail with a generic “no base image” error with the stacktrace containing an error stating that the previous test base screenshot was not present:

First test failure: 

Minitest::Assertion:
  Screenshot does not match for 'features-email_notification_preferences_spec/email_preferences_with__review_requests_should_allow_rescan_customers_to_view_rescan_preference/00_rescan_preference_page_visited' ({"area_size":263590.0,"region":[256.0,19.0,1482.0,234.0],"difference_level":0.006523589743589743})
  /home/circleci/project/spec/screenshots/linux/features-email_notification_preferences_spec/email_preferences_with__review_requests_should_allow_rescan_customers_to_view_rescan_preference/00_rescan_preference_page_visited.png
  /home/circleci/project/spec/screenshots/linux/features-email_notification_preferences_spec/email_preferences_with__review_requests_should_allow_rescan_customers_to_view_rescan_preference/00_rescan_preference_page_visited.base.diff.png
  /home/circleci/project/spec/screenshots/linux/features-email_notification_preferences_spec/email_preferences_with__review_requests_should_allow_rescan_customers_to_view_rescan_preference/00_rescan_preference_page_visited.diff.png
/home/circleci/project/vendor/bundle/ruby/3.1.0/gems/rspec-core-3.12.0/lib/rspec/core/example.rb:457:in `instance_exec'......

Second test failure:
Failure/Error: raise "There is no original (base) screenshot version to compare, located: #{@base_image_path}" unless @base_image_path.exist?

RuntimeError:
  There is no original (base) screenshot version to compare, located: /home/circleci/project/spec/screenshots/linux/features-email_notification_preferences_spec/email_preferences_with__review_requests_should_allow_rescan_customers_to_view_rescan_preference/00_rescan_preference_page_visited.base.png
/home/circleci/project/vendor/bundle/ruby/3.1.0/gems/capybara-screenshot-diff-1.8.0/lib/capybara/screenshot/diff/image_compare.rb:106:in `_different?'
/home/circleci/project/vendor/bundle/ruby/3.1.0/gems/capybara-screenshot-diff-1.8.0/lib/capybara/screenshot/diff/image_compare.rb:67:in `different?'
/home/circleci/project/vendor/bundle/ruby/3.1.0/gems/capybara-screenshot-diff-1.8.0/lib/capybara/screenshot/diff/test_methods.rb:92:in `assert_image_not_changed'
/home/circleci/project/vendor/bundle/ruby/3.1.0/gems/capybara-screenshot-diff-1.8.0/lib/capybara/screenshot/diff.rb:95:in `block in track_failures'
/home/circleci/project/vendor/bundle/ruby/3.1.0/gems/capybara-screenshot-diff-1.8.0/lib/capybara/screenshot/diff.rb:94:in `map'
/home/circleci/project/vendor/bundle/ruby/3.1.0/gems/capybara-screenshot-diff-1.8.0/lib/capybara/screenshot/diff.rb:94:in `track_failures'
/home/circleci/project/vendor/bundle/ruby/3.1.0/gems/capybara-screenshot-diff-1.8.0/lib/capybara/screenshot/diff.rb:83:in `block in included'....

This bug was caused by the fact that in diff.rb the tested screenshots are cleared out after track_failures. This is a problem because track_failures raises an error if the screenshots are different; causing the code to break and not return to the clearing of the testing screenshots. In this PR, I have moved the clearing of the testing screenshots to happen inside of track_failures after the errors have been collected from the test screenshots.

[pftg]

@TaylorBrysonRouse, nice catch. Thanks!