Consider all domain suffixes for blacklisting and normalise implementations

[owst]

This change was triggered by the Ruby implementation not matching any blacklisted domains with more than 1 subdomain, e.g. test@spamthis.co.uk would not be considered invalid, even though spamthis.co.uk is a blacklisted domain.

This refactoring normalises the implementation across the 7 languages:

1. We're now using the same regexp for email wellformed-checking,
2. All suffixes of an email's domain are checked for blacklist membership in the same way (modulo language/library differences),
3. The test cases are now the same for each implementation.

This removes false-negatives (e.g. where the domain includes > 2 subdomains) in some cases, and false-positives (e.g. where a blacklisted domain appears as a subdomain) in other cases.

As a side effect, several improvements were made to the tests of PHP, Python in particular, using standard libraries for unit testing rather than ad-hoc tests.

Meta: I think this will be easiest to review on a commit-by-commit basis. This PR is based off my other pending PR, so that PR's commit also shows up here.

[owst]

I can't work out why circle ci is failing on composer install --prefer-source --no-interaction I was able to run that locally with no problem. Perhaps a version issue, I'm on:

~ composer --version
Composer version 1.0.0-alpha11 2015-11-14 16:21:07

[FGRibreau]

Very impressive work @owst!! Thanks a lot!

[owst]

You're welcome, thanks for maintaining the project!

How do we go about getting a new Ruby gem published? At work we're currently overriding MailChecker.valid_email?, so it'd be good to get rid of that code by using a newly published version.

By the way, do you have any idea why composer failed on Circle CI?

[FGRibreau]

Hello @owst ,

It's now fixed, I pushed ruby-mailchecker-1.6.3.gem :)

[owst]

Great, thanks!