Fix candidate `index` for new -002 models and enable `candidateCount`

For unknown reason 0-index is absent in new -002 models response
Ref: https://ai.google.dev/api/generate-content#v1beta.Candidate

Note: atm `candidateCount` works only for -002 models
      (but not for `streamGenerateContent` endpoint)

fixup! Allow specifying arbitrary "gemini-*" model

Remove unneeded code

`gemini-1.5-pro-latest` successfully accept multiple user messages,
so it's now no need to insert dummy model messages between them