Consider switching from the `text` package to `text-utf8` or the like for the high-level api.

[zenhack]

...since capnp itself uses utf8 encoding internally. This would likely improve performance.

[zenhack]

Note that text 2.0 makes this change as well; we should just upgrade the package (and then make use of the new encoding).

[zenhack]

I've updated the version bounds so we now support text 2.0. Before closing this though, we should make the Parse implementation for Basics.Text zero-copy.