Improve reproducibility with faker

[nikophil]

When using Faker, I prefer to have full randomness from one execution to another (ie: not specifying a seed)

but, when a test fails, I would like to reproduce the problem, with the exact same data, which is not actually possible, because it used an unknown seed.

My proposal would be to always give a random seed to faker, and then display it right after phpunit's execution (thanks to our PHPUnit extension). And being able to set the seed thanks to an environment variable.

I think this would also mean to deprecate zenstruck_foudry.faker.seed configuration.