fixing mypy's Matcher[Never]

[bukzor]

The idea is that manually unrolling these Union[Matcher[T], T] alternatives into @overload alternatives allows mypy to properly evaluate them in order, rather than getting confused and giving up (resulting in an inference of T = Never).

Closes: #266 #264

Two off-topic changes included: (you can revert them or ask me to)

• uv sync will now treat PyHamcrest[dev] as its dev "dependency group"
• full typing for the implementation of has_entries

[bukzor]

draft: adding some quick typechecking tests

[brunns]

Looks like we'd need to remove 3.8 support to make this work, which I am fine with.

[bukzor]

3.8 compat restored. All type annotations regression tested (they fail with the "Matcher[Never]" error at HEAD, mostly).

[bukzor]

Ready for merge, to the best of my ability.

[bukzor]

happy monday!