fix(openai): correct token details field names for Response API usage (#1564)

fix(openai): correct token details field names for Response API usage parsing

The _parse_usage() function checked for input_token_details and
output_token_details (singular), but OpenAI's Response API returns
input_tokens_details and output_tokens_details (plural). This caused
nested token detail fields (cached_tokens, reasoning_tokens) to be
silently ignored.

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Hassieb Pakzad <68423100+hassiebp@users.noreply.github.com>