min_size arguments should default to zero instead of None

[Zac-HD]

We default arguments to None where we don't have a known bound (e.g. integers()), or where a non-None default argument would require us to weaken our checks for consistency (e.g. if min_value defaulted to -inf for floats(), we couldn't warn a user who passed min_value=-inf, allow_infinity=False). For strategies such as dates() we don't accept None at all!

It turns out that neither of these reasons apply to minimum sizes! We should therefore:

• Set min_size=0 as the default for all size arguments
• In hypothesis.internal.validation.check_valid_size, add a check if name == 'min_size' and value is None: and use note_deprecation to warn users that they should use min_size=0 instead.
• Search for any remaining cases in our codebase where we use this argument, with git grep --line-number --perl-regexp "min_size ?= ?None"
• Write a test in hypothesis-python/tests/cover/test_validation.py that uses @checks_deprecated_behaviour and passes min_size=None to lists().

[lgoeller]

I'm going to take this one on!