When decoding a string in BSON, the driver attempts to validate that the string is valid UTF-8, which can cause issues if the data in the database is actually invalid. We should allow users to read out that data, despite it being invalid, because Go strings are just bytes and don't conform to a specific charset.

The existing check is also incorrect because it's only checking single-byte strings, but multi-byte strings can be invalid UTF-8 as well.