meta generated for messages with oneOf properties is incorret for other properties

[jkristia]

I see an issue in the generated meta when a class contains a oneof type.
this class

message UsingOneOf {
  type1.Type1Basic some_type = 1;
  oneof one_of_type {
    type1.Type1Basic one_of_type1 = 3;
    type2.Type2Basic one_of_type2 = 4;
  }
}

generates oneOfIndex: 0 for all 3 properties, not just the 2 properties in the oneof field.
This is with version

    "ts-proto": "^1.170.0",
    "ts-proto-descriptors": "^1.15.0",

The generated meta. Notice field some_type also has oneOfIndex: 0

   {
      "name": "UsingOneOf",
      "field": [{
        "name": "some_type",
        "number": 1,
        "label": 1,
        "type": 11,
        "typeName": ".type1.Type1Basic",
        "extendee": "",
        "defaultValue": "",
        "oneofIndex": 0,   <----
        "jsonName": "someType",
        "options": undefined,
        "proto3Optional": false,
      }, {
        "name": "one_of_type1",
        "number": 3,
        "label": 1,
        "type": 11,
        "typeName": ".type1.Type1Basic",
        "extendee": "",
        "defaultValue": "",
        "oneofIndex": 0,
        "jsonName": "oneOfType1",
        "options": undefined,
        "proto3Optional": false,
      }, {
        "name": "one_of_type2",
        "number": 4,
        "label": 1,
        "type": 11,
        "typeName": ".type2.Type2Basic",
        "extendee": "",
        "defaultValue": "",
        "oneofIndex": 0,
        "jsonName": "oneOfType2",
        "options": undefined,
        "proto3Optional": false,
      }],
      "extension": [],
      "nestedType": [],
      "enumType": [],
      "extensionRange": [],
      "oneofDecl": [{ "name": "one_of_type", "options": undefined }],
      "options": undefined,
      "reservedRange": [],
      "reservedName": [],
    }

[stephenh]

@lukealvoeiro I haven't looked into this very much, but oneofIndex is a proto2 field; could this be from a ts-proto-descriptor thing / your recent PR?

You'd mentioned bumping ts-proto-descriptors, but I don't think that ended up happening, but I kinda thought it shouldn't really matter since it'd still be using the generated-by-older-ts-proto versions of the decode code... 🤔