Incorrect enforcement of the `length` trait when applied to string shapes

[david-perez]

As far as I can see, the code is enforcing Smithy's length constraint trait as a call to the JavaScript's object length property.

This implementation is wrong for JavaScript Strings, because its length property counts the number of UTF-16 code units, but the Smithy spec says length counts the number of Unicode code points when applied to strings.

These are not the same; consider characters outside the basic multilingual plane. U+10000 is encoded using two UTF-16 code units, but is only one Unicode code point (scalar value).

This would have been caught with Smithy protocol tests: smithy-lang/smithy#1090.