Previously we were inconsistent by generating code for enum definitions
but not generating code for dictionaries. With future changes to the
IDL generator to expose helpers to convert to and from IDL values
this produced circular depdendencies. To solve this problem, also
generate the dictionary definitions in bindings headers.
This is the streaming counterpart to TextDecoder, used by sites that
process responses with `pipeThrough(new TextDecoderStream())` or that
otherwise consume a Response body as decoded text.
The transform algorithm appends each chunk's bytes to the I/O queue
and decodes them via the underlying LibTextCodec decoder, holding back
any trailing partial UTF-8 sequence so chunk boundaries don't produce
spurious replacement characters. The flush algorithm emits a single
replacement character if a partial sequence was still pending. For
non-UTF-8 encodings the underlying decoders are stateless across
calls, so each chunk is decoded in full with nothing carried over.
