Applying `to_iri` on two `URL`s can make change the equality of URLs.

```python
from hyperlink import URL
url1 = URL.from_text(u"http://example.test./,")
url2 = URL.from_text(u"http://example.test./%2c")
assert url1 != url2
assert url1.to_iri() == url2.to_iri()
```
It seems surprising to me that `to_iri()` provides a level of normalization that is not available via `.normalize()`

This seems related to #18 and #164.