[tcgc] should tcgc unify the content type logic for bytes format inference

[qiaozha]

for this case, the contentType is a union of application/json or text, but tcgc returns bytes as the default encoding

  @get
  getSchemaByVersion is Azure.Core.Foundations.Operation<
    Azure.Core.Foundations.ItemKeysOf<Version>,
    OkResponse & SchemaContent & SchemaContentType
  >;

@doc("The content type for given schema. Each schema type has an associated content-type.")
model SchemaContentType {
  @doc("The content type for given schema.")
  @header("Content-Type")
  contentType: SchemaContentTypeValues;
}

@doc("Content of the schema.")
model SchemaContent {
  @body
  @doc("String representation (UTF-8) of the schema.")
  content: bytes;
}

union SchemaContentTypeValues {
  avro: "application/json; serialization=Avro",
  @added(ServiceApiVersions.V2022_10)
  json: "application/json; serialization=json",
  @added(ServiceApiVersions.V2022_10)
  custom: "text/plain; charset=utf-8",
  @added(ServiceApiVersions.V2023_07_01)
  protobuf: "text/vnd.ms.protobuf",
}

After some investigation with @tadelesh offline, tcgc is using strict equal to "application/json" to set the default encoding as base64

typespec-azure/packages/typespec-client-generator-core/src/types.ts

Line 184 in 5506281

} else if (!defaultContentType || defaultContentType === "application/json") {

But, previously, we have an ask from client emitter side that suggests us to leverage the logic from dotnet Azure/autorest.typescript#2200 and fixed in https://github.com/Azure/autorest.typescript/pull/2258/files#diff-7dd8f9407e7839a21c48e98c9da493fa7da52885017e1df4dbf18a346b9933c8

I wonder if tcgc could also leverage this ?