* capybara spec changes

Fixed:

> Disable compile cache if Ruby 3.0.3's ISeq cache bug is detected. AKA iseq.rb:13 to_binary: wrong argument type false (expected Symbol)

Previously we assumed that `db/schema.rb` is database agnostic.
This is not exactly true in all regards.

* rails/rails#26209 (comment)

Our commited `schema.rb` can be used to `db:setup` all databases.
running `db:schema:dump` or `db:migrate`, will turn the file into a
database specific one.
Formerly that wasn't obvious with the sqlite adapter, but now it dumps
foreign key information into the schema.
The previous check cannot be used anymore.

This is unfortunate, since frab's migrations are so old, that you cannot
`db:create db:migrate` a fresh system.

Not sure if related, but we will probably see more problems with old
frab installations on MySQL, due to the 2016 bigint change. These become
especially problematic, when using foreign key constraints:

* (MySQL switched to bigint)[rails/rails#26266]
* rails/rails#43168

Changing all IDs and references to bigint on MySQL is probably not
possible from migrations and should be done by the admin. A more
extensive solution than this is needed:
#660 (comment)

* this drops database_cleaner, the default transactional tests work fine
* don't fail-fast the github actions matrix, so we now if it's a
  database related failure
* frab test event title doesn't contain database id, multiply by 3 to
  avoid flakes in filter tests
* editing_event_people_test waits for hidden field to change
* turn on parallelized tests, one can set PARALLEL_WORKERS for more
  workers

The quotes are needed, because YAML parses the value to a number.
This turns the 3.0 into a 3, which leads to the installation of Ruby
3.1.
Howver gems (github-markdown?) are not ready for 3.1