spec_helper: respect existing 'LOCAL_IP_ADDR' env var

See merge request !406

specs: relax expected error message if refreshing keys fails

See merge request !407

This change allows to set the env var before building the Debian
package, which otherwise fails, as the related code doesn't work if
invoked in a chroot, for unknown reasons.

As of 2022/04/16, gnupg 2.2.27-3 as shipped in Debian, reports 'unknown
host' and 'connection refused', which is different from what we see in
our upstream CI. Again, reasons are unclear.

Given this, relax the expected error message.

Run test suite with Ruby 3.1

Closes #511

See merge request !396

That class was removed from the stdlib and extracted into a gem. The
`mail` didn't yet release a new version to include that gem as a
dependency itself.

specs: make GPG key dates strict, again

See merge request !404

We decided to enforce a specific timezone and locale to be used within
the specs, see !402 for details. Doing so makes working with the
testsuite easier for humans, instead of relying on regexes to make
machines happy.

CI: rspec: display doc strings of specs to ease insight

See merge request !405

CI: drop job to check for GPG key dates, again

See merge request !403

Force a specific timezone and locale

Closes #515

See merge request !402

We decided to instead enforce a specific timezone and locale to be used
within the specs, see !402 for details.

This is required so tests can sucessfully match on dates and strings
regardless of the timezone and locale settings of the system they run
on.

CI: Add job to check for expected, hardcoded key expiry dates in specs

Closes #514

See merge request !400

specs: key_management: relax expected key expiry dates

Closes #513

See merge request !399

specs: make it possible to conditionally require 'byebug'

See merge request !398

Specs which expect such dates will fail during reproducible builds, as
such builds rely on timezone-variations, and these dates are, in fact,
timezone-sensitive.

Closes #514

The expiry dates of GPG keys are timezone-sensitive. To make Schleuder
build reproducible, relax these dates.

Closes #513
Ref #268
Ref !118

Add fix for bounces_drop_all bug

See merge request !392

This helps in downstream environments, like Debian, via relying on
environment variables, instead of patches, to not require 'byebug'.

Similar code is already in place in regards to the usage of 'bundler'
and 'simplecov'.

This commit doesn't include a CHANGELOG note, as it's not really
relevant to most users, I believe.

CI: changelog: enforce latest state

Closes #512

See merge request !397

When bounces_drop_all=true, filters such as receive_encrypted_only were
circumvented and allowed non-encrypted senders to pass messages onto the
lists (see #508).

This moves the bounce-dropping mechanism from `filters_runner.rb` to
`runner.rb`. Thus, the decision if a bounce is being dropped will be made
entirely in runner.rb. `filters_runner.rb` will only take care of checking
for errors.

This might be necessary to handle non-fast-forward changes, which would
be rejected otherwise.

Closes #512

Use local IP address for sks-mock, not localhost

Closes #509 and #478

See merge request !394

If using the "standard-resolver" and a not reachable keyserver, dirmngr
reports a different error message than with its internal resolver 🤦

"localhost" is not available in (most?) docker environments.

This also makes dirmngr disable its internal resolver, which apparently
has problems with HKP-URLs containing local IPs.

CHANGELOG: mention improvements of the German locales

See merge request !391

Check stderr and exitcode in CLI specs

See merge request !393

Cli#install runs migrations which might cause errors and a SystemExit.
Unless we catch and maybe check for those they can result in the whole
test runner exiting and thus the suite being short-cut.

db: add migration to change boolean values to integers

Closes #505

See merge request !389

It seems, the change to ActiveRecord >= 6.0 makes this necessary.
Without doing this, the auto-generated database schema file would drop
these limits, if running migrations.

It seems, this problems has been ongoing since quite some time. For
details, see [1].

[1] https://github.com/rails/rails/issues/19001