Describe the issue or request
The latest directive from azure is to support this: https://github.com/microsoft/api-guidelines/blob/685e493d38f8a3875c22336dcd177f4b54dcfb23/azure/Guidelines.md#returning-string-offsets--lengths-substrings

We've implemented the model in the pr here, but the idea from the API Review team is that the SDK's would automatically select the string encoding the language supports and ignore the other two, thereby exposing this field as an int instead of a model.

Azure/azure-sdk-for-net#43913

Within the code:

// Currently
taggerOutput.Offset.Utf16 // int
taggerOutput.Offset //  {"utf8": 10, "utf16":10, "codePoint:10"} - StringIndex model

// Desired
taggerOutput.Offset // int -- While still supporting the serialization of the StringIndex model