Do not use Vec<u8> as argument type in the binding

[vxgmichel]

See the corresponding issue on pyo3 for more information: PyO3/pyo3#2888

Instead we should take a PyObject as argument and perform an explicit conversion:

fn extract_to_vec_u8(object: PyObject, py: Python) -> PyResult<Vec<u8>> {
    Ok(match object.extract::<&PyByteArray>(py) {
        Ok(x) => x.to_vec(),
        Err(_) => object.extract::<&PyBytes>(py)?.as_bytes().to_vec(),
    })
}

Note: this performs a copy than can usually be avoided but it requires an unsafe block in the case of a bytearray. There's already a couple of such blocks in the codebase, e.g:

parsec-cloud/oxidation/libparsec_python/src/api_crypto.rs

Lines 29 to 35 in 9fbd8ef

	let bytes = match data.extract::<&PyByteArray>(py) {
	// Using PyByteArray::as_bytes is safe as long as the corresponding memory is not modified.
	// Here, the GIL is held during the entire access to `bytes` so there is no risk of another
	// python thread modifying the bytearray behind our back.
	Ok(x) => unsafe { x.as_bytes() },
	Err(_) => data.extract::<&PyBytes>(py)?.as_bytes(),
	};