Add zero-copy PyString wrapper

[cjdoris]

Motivation

• Provide a zero-copy wrapper for Python str values so Julia code can access UTF-8 bytes without allocating or copying Python data.
• Reuse and generalize the existing UTF-8 string iteration/validation logic so both the new wrapper and StaticString share correct behaviour.

Description

• Add exported PyString <: AbstractString (stores the underlying Py object, a Ptr{UInt8} to the UTF-8 data and its byte length) in src/API/types.jl and implement the wrapper in src/Wrap/PyString.jl.
• Expose the C API binding PyUnicode_AsUTF8AndSize in src/C/pointers.jl and use it to construct PyString with a pointer+length without copying.
• Generalise the UTF-8 helpers in src/Utils/Utils.jl by introducing utf8_isvalid and utf8_iterate and adapt StaticString helpers to reuse them, and add bounds-checked codeunit for PyString.
• Register a normal-priority pyconvert rule mapping builtins:str → PyString in src/Wrap/Wrap.jl so PyString conversion is available while preserving the existing canonical String conversion rules; export PyString via src/API/exports.jl.
• Add tests exercising ASCII, multi-byte Unicode, embedded NULs, empty strings, bounds, invalid-UTF8 detection, iteration and conversion priority in test/Convert.jl and test/Utils.jl.

Testing

• Ran package tests with julia --project -e 'using Pkg; Pkg.test()' and the test suite completed with PythonCall tests passing (test summary shows package tests passed while two pre-existing tests were marked broken in the environment).
• Ran Python tests with cp pysrc/juliacall/juliapkg-dev.json pysrc/juliacall/juliapkg.json && uv run pytest -s --nbval ./pytest and all Python tests passed (21 passed).
• Ran coverage with julia --project -e 'using Pkg; Pkg.test(coverage=true)' and the coverage run completed; the new PyString implementation and the edited UTF-8 utility code have executable lines covered by tests.
• Added unit tests in test/Convert.jl and test/Utils.jl that passed during the above runs.

---

Codex Task