[PGZXB]

Fixes #336.

lol_html_attributes_iterator_get returned a std::slice::Iter borrowed from the element's attribute Vec. lol_html_element_set_attribute (and lol_html_element_remove_attribute) grow or rebuild that Vec, so the next call to lol_html_attributes_iterator_next walked a stale slice cursor into a buffer that had already been freed, even though the iterator is documented to stay valid for the whole handler.

The iterator now carries the element pointer and a cursor and _next re-reads the element's attributes on each call, so mutating the element between calls no longer invalidates the iteration:

lol_html_attribute_t *attr = lol_html_attributes_iterator_next(iter);   // first attribute
lol_html_element_set_attribute(element, "zzzzzzzzzzzzzzzzzzzz", 20, "1", 1);
attr = lol_html_attributes_iterator_next(iter);                         // next attribute, no UAF

Verified with the reproducer from the issue (<div a="1" b="2" c="3">, advance once, mutate, advance again), built against the C API with AddressSanitizer. Before the change it reports ERROR: AddressSanitizer: heap-use-after-free on the second next; after the change it runs cleanly and prints the next attribute. cargo fmt --check and cargo clippy --manifest-path=c-api/Cargo.toml are clean.