[Serde generate] runtimes: negative fuzzing + UTF8 validation + max container depth

[ma2bd]

Summary

• Use Rust as a baseline to test that supported languages reject incorrect byte strings during LCS deserialization
• Introduce a generated list of "negative" samples for this purpose. Samples are typically obtained by flipping the highest bit of a byte in a valid sample, and testing if Rust can deserialize.
• In all languages, we now also mutate "positive samples" and check that they always decode to a different value (by canonicity)
• The new testing approach covers exceeding input bytes, UTF8 validation, boolean decoding, map key ordering, the length limit of sequences, and the limit on container depth.
• Also, fix a codegen issue in golang for containers of the form struct Foo(Option<T>)

Test Plan

CI