this commit fixes bad runtime encoding of unicode strings. `parse_Str`
used the utility function `string_to_bytes`, which rejects characters
with value larger than 255 and otherwise produces the ascii encoding
of the string. the issue is that bytes in the range 128-255 specify
different characters in utf-8 than in ascii encodings, resulting in
different values at runtime than at compile-time.

this can be seen from differing compile-vs-runtime behavior of
`keccak256` (this example was provided in GH issue 3088):

```vyper
@external @view def compile_hash() -> bytes32:
    return keccak256("è")

@external @view def runtime_hash() -> bytes32:
    s: String[1] = "è" return keccak256(s)
```

this commit fixes and simplifies `parse_Str` by using python's
`str.encode()` builtin, which encodes using utf-8 by default. it also
increases strictness of string validation to reject bytes in the range
128-255, since in utf-8 these can encode multibyte characters, which we
reject in vyper (see more discussion in GH issue 2338).