st.from_regex() alphabet improvements

[Zac-HD]

Ensure that unions of st.characters() and st.sample_from() are usable as alphabet= arguments in st.text() and st.from_regex().

We don't want to auto-convert all such things to characters() by default because that affects shrink order, but as the alphabet argument where we're already more constrained it's fine. Motivating example based on zarr-developers/zarr-python#1746:

from hypothesis import given, strategies as st

zarr_key_chars = st.sampled_from("abc")
array_names_filter = st.text(zarr_key_chars | st.just("."), min_size=1).filter(lambda t: t[0] != ".")
array_names_regex = st.from_regex(
    r"[^.].*",
    fullmatch=True,
    alphabet=zarr_key_chars | st.just("."),
    # InvalidArgument: alphabet=one_of(sampled_from('abc'), just('.')) must be a sampled_from() or characters() strategy
)

@given(array_names_regex)
def test(name): ...

---

Separately, we should remove the prints here, and change "prefix caret" to "prefix caret to group containing pattern"; in the current form the semantics of match a|b is converted to match a|.*b.

hypothesis/hypothesis-python/src/hypothesis/strategies/_internal/strings.py

Lines 192 to 198 in 395649a

	print(f"{condition=}")
	print(f"{condition.__name__=}")
	if condition.__name__ == "match":
	# Replace with an easier-to-handle equivalent condition
	caret = "^" if kind is str else b"^"
	pattern = re.compile(caret + pattern.pattern, flags=pattern.flags)