Fix FuzzTests unit tests

[gjcairo]

One of the assertions in the TextFormat fuzz-originated unit tests checked that we failed to decode a string built from a sequence of bytes. However, this sequence of bytes had invalid UTF8 characters. Foundation has improved in recent betas and now fails to build a String from this sequence (which it didn't before). This is causing the test to fail.

If the String couldn't have been constructed during fuzz testing, the test would have passed - see:

swift-protobuf/FuzzTesting/Sources/FuzzTextFormat/main.swift

Line 20 in 601e03a

guard let str = String(data: Data(bytes), encoding: .utf8) else { return 0 }

We should tolerate the String not being constructed here without failing the test, as it's the correct (and current) behaviour. This means that after this PR, assertions will be skipped when a String cannot be constructed.
However, we shouldn't simply remove the failing assertion, because in the case Foundation introduces this bug again, we should make sure we're failing the TextFormat decoding.

[thomasvl]

Fyi - #1590 is the PR that removed the original test case. We'd pulled it in to be a unittest to help keep a better eye on the handling.